Prompts
Agent Planning and Task Decomposition
Task Decomposition and Subtask Generation
Prompt playbooks for breaking a complex objective into a structured set of smaller, executable subtasks. Useful for agent orchestrators and autonomous workflow builders because decomposition quality directly determines whether downstream planning, tool selection, and execution steps remain coherent and completable.
Complex Objective to Subtask Decomposition Prompt Template
For agent orchestrators that must convert a high-level user goal into a structured, executable subtask list. Produces a JSON task graph with dependencies, tool assignments, and completion criteria. Includes eval checks for task atomicity, coverage of the original objective, and dependency cycle detection.
Goal-to-Task Breakdown Prompt with Tool Constraints
For planning modules that must decompose a goal while respecting a known, limited tool set. The output is a feasible task sequence that only uses available tools, with explicit flags for steps that require unavailable capabilities. Harness must validate that no hallucinated tools appear in the plan.
User Intent to Executable Steps Decomposition Prompt
For agent frontends that translate ambiguous user requests into a concrete step-by-step plan. Produces a clarified goal statement, ordered steps, and explicit clarification questions for ambiguous parts. Evaluation criteria include intent preservation and appropriate abstention on under-specified requests.
Hierarchical Task Network Generation Prompt Template
For workflow builders that need a recursive task decomposition from a compound goal down to primitive actions. Outputs a tree structure with parent-child relationships, execution constraints, and leaf-node tool bindings. Tests must check for valid hierarchy, complete decomposition to primitives, and no orphaned subtasks.
Dependency-Aware Task Decomposition Prompt
For agent architects who need a task list with explicit prerequisite relationships and a valid topological ordering. Produces a directed acyclic graph of subtasks with dependency edges and parallel execution groups. Harness must detect cycles, missing dependencies, and tasks that depend on outputs from later steps.
Parallel Subtask Identification Prompt for Agent Workflows
For orchestrators that need to identify which subtasks can execute concurrently to minimize end-to-end latency. Outputs task groups annotated with parallel-safe markers and shared-resource conflict warnings. Evaluation checks for false parallelism that would cause race conditions on mutable state.
Codebase Feature to Subtask Breakdown Prompt
For coding agents that must decompose a feature request into implementation subtasks within a specific repository context. Produces file-level change descriptions, ordering constraints, and test requirements. Harness must validate that subtasks reference real file paths and respect module boundaries.
Bug Fix Decomposition Prompt for Coding Agents
For coding agents that receive a bug report and must produce a diagnostic and fix plan. Outputs root-cause investigation steps, candidate fix locations, verification subtasks, and regression test requirements. Evaluation criteria include whether the plan isolates the bug before attempting fixes.
Incident Response Subtask Generation Prompt
For on-call automation agents that must decompose an incident alert into triage, mitigation, investigation, and communication subtasks. Produces a time-ordered runbook with severity-appropriate actions and escalation gates. Harness must check that irreversible actions are gated behind confirmation steps.
Data Pipeline Task Decomposition Prompt
For data engineering agents that must break a pipeline objective into extraction, validation, transformation, and loading subtasks. Outputs a staged pipeline plan with schema contracts between stages and error-handling branches. Tests validate data lineage traceability across the decomposition.
RAG System Build Subtask Generation Prompt
For AI engineers decomposing a retrieval-augmented generation system build into chunking, embedding, indexing, retrieval, and generation subtasks. Produces a build plan with component interfaces, evaluation checkpoints, and configuration decisions. Harness must flag missing evaluation steps in the plan.
API Integration Task Decomposition Prompt
For integration agents that must break an API connection objective into discovery, authentication, schema mapping, error handling, and testing subtasks. Outputs a sequenced integration plan with endpoint-specific steps and fallback logic. Evaluation checks for missing auth or rate-limit handling.
Security Audit Task Breakdown Prompt
For security review agents that must decompose an audit scope into reconnaissance, vulnerability assessment, finding documentation, and remediation planning subtasks. Produces an audit work plan with evidence collection steps and risk-prioritized task ordering. Harness must validate that destructive tests are flagged for approval.
Customer Support Ticket Resolution Task Breakdown Prompt
For support automation agents that must decompose a ticket into information gathering, diagnosis, resolution, and verification subtasks. Outputs a resolution plan with customer-facing steps, internal lookup steps, and handoff criteria. Evaluation checks for inappropriate automation of sensitive or high-risk tickets.
Research Synthesis Task Decomposition Prompt
For research agents that must break a synthesis objective into source collection, claim extraction, contradiction analysis, and synthesis writing subtasks. Produces a research plan with evidence-grounded steps and uncertainty annotations. Harness must validate that synthesis steps are gated on evidence collection completion.
Product Requirements Decomposition Prompt
For product agents that must decompose a product brief into user stories, acceptance criteria, technical constraints, and milestone groupings. Outputs a structured requirements breakdown with dependency-aware ordering. Tests check for traceability from each subtask back to the original product objective.
Browser Automation Task Decomposition Prompt
For computer-use agents that must decompose a web task into navigation, element interaction, data extraction, and state verification subtasks. Produces a browser action plan with wait conditions and error recovery branches. Harness must validate that each step includes a verifiable success condition.
SOP Automation Task Breakdown Prompt
For workflow automation agents that must convert a standard operating procedure document into executable automation subtasks. Outputs a structured task list with decision points, tool mappings, and human approval gates. Evaluation criteria include preservation of all SOP branches and correct identification of steps that cannot be automated.
Multi-Agent Handoff Subtask Generation Prompt
For multi-agent orchestrators that must decompose a goal into subtasks assigned to specialized sub-agents with defined handoff contracts. Produces an agent assignment map with input/output schemas per handoff and coordination checkpoints. Harness must detect missing context in handoff payloads.
Human-in-the-Loop Approval Task Decomposition Prompt
For supervised autonomy systems that must decompose a workflow into AI-executable steps and human-review gates. Outputs a task sequence with explicit approval checkpoints, review criteria, and timeout/fallback behavior. Tests validate that irreversible actions are always gated and that the plan degrades safely if approval times out.
RAG Query Decomposition Prompt for Complex Questions
For retrieval systems that must decompose a multi-hop question into sub-questions answerable from separate document chunks. Produces a set of atomic retrieval queries with logical operators and evidence combination rules. Harness must check that sub-questions are independently answerable and that the combination logic is sound.
Model Fine-Tuning Task Breakdown Prompt
For ML engineers decomposing a fine-tuning project into data curation, formatting, training configuration, evaluation design, and deployment subtasks. Produces a project plan with quality gates between stages and rollback criteria. Evaluation checks for missing data contamination safeguards.
Prompt Engineering Workflow Decomposition Prompt
For prompt engineers decomposing a prompt development cycle into requirement gathering, draft generation, test case creation, evaluation, and iteration subtasks. Produces a development plan with eval harness integration points and versioning checkpoints. Harness must flag plans that skip systematic evaluation.
Threat Modeling Task Decomposition Prompt
For security agents that must decompose a system threat modeling exercise into asset identification, threat enumeration, control analysis, and risk rating subtasks. Produces a structured assessment plan with evidence requirements per threat. Tests validate that the decomposition covers all STRIDE categories when applicable.
Failure Mode Analysis Subtask Generation Prompt
For reliability agents that must decompose a failure mode analysis into component identification, failure scenario generation, impact assessment, and mitigation planning subtasks. Outputs an FMEA-style work plan with severity and likelihood rating steps. Harness must check that detection mechanisms are included for each failure mode.
Dependency Mapping and Execution Order
Prompt playbooks for identifying prerequisite relationships between tasks and determining a valid execution sequence. Useful for workflow engine developers and agent architects because incorrect ordering causes deadlocks, wasted tool calls, and partial completions that are expensive to unwind.
Task Dependency Graph Extraction Prompt Template
For workflow engine developers and agent architects. Extracts a structured dependency graph (nodes, edges, prerequisite relationships) from a natural language task description. Includes eval checks for missing transitive dependencies and orphan tasks.
Execution Order Linearization Prompt Template
For agent orchestrators that need a valid linear execution sequence from a dependency graph. Produces a topologically sorted task list with parallel execution groups identified. Harness must validate that no task appears before its prerequisites.
Circular Dependency Detection Prompt
For workflow validation pipelines. Analyzes a set of task definitions and their declared prerequisites to identify cycles that would cause deadlocks. Output includes the cycle path and a recommended break point. Requires human review before automated resolution.
Parallel Execution Grouping Prompt Template
For agent runtimes optimizing throughput. Groups independent tasks from a dependency graph into parallel execution waves while respecting resource constraints. Evaluation checks that no grouped tasks share a dependency path.
Missing Dependency Discovery Prompt
For plan validation before execution. Reviews a task plan and identifies tasks that reference outputs, resources, or state from undeclared or incomplete upstream steps. Output is a remediation list with suggested prerequisite additions.
Deadlock Detection Prompt for Workflow Definitions
For CI/CD and workflow engine operators. Analyzes a formal workflow definition (YAML, JSON, or DSL) to detect resource-wait cycles, mutex conflicts, and ordering deadlocks. Includes severity classification and resolution guidance.
Critical Path Analysis Prompt for Task Sequences
For project-aware agent planners. Identifies the critical path through a dependency graph with duration estimates, highlighting tasks that directly impact total completion time. Output includes slack time for non-critical tasks.
Agent Handoff Dependency Contract Prompt Template
For multi-agent system architects. Defines the input/output contract between two agents in a dependency chain, including required fields, expected state, failure modes, and timeout behavior. Used to prevent silent context corruption at handoff boundaries.
Context Passing Dependency Verification Prompt
For chained agent pipelines. Verifies that the output from an upstream agent contains all required fields and sufficient context for the downstream agent to proceed. Flags missing, stale, or ambiguous data before the dependent step executes.
Retry Dependency Cascade Analysis Prompt
For resilient agent runtimes. When a task fails and requires retry, analyzes which downstream tasks are invalidated and must be re-executed versus which can safely proceed. Prevents partial-state corruption from selective retries.
Side Effect Ordering Constraint Prompt Template
For agents operating on mutable external resources. Identifies ordering constraints driven by side effects (database writes, API mutations, file changes) that are not captured by explicit data dependencies. Includes rollback safety analysis.
Rollback Safe Point Identification Prompt
For workflow designers building reversible pipelines. Analyzes a task sequence and identifies safe rollback points where state can be cleanly reverted without cascading side effects. Output is a checkpoint map for the execution engine.
Resource Contention Prediction Prompt
For agent schedulers managing limited tool or API capacity. Predicts which tasks in a dependency graph will contend for the same resource and recommends staggering or serialization. Includes rate limit and concurrency constraint awareness.
API Dependency Mapping Prompt for Integration Plans
For integration engineers building multi-service workflows. Extracts API call dependencies from an integration plan, identifying which endpoints must be called before others based on authentication, resource creation, or data flow requirements.
Deployment Order Sequencing Prompt Template
For DevOps and release engineers. Generates a validated deployment sequence for multiple services, databases, and configuration changes based on declared dependencies. Flags missing dependencies that could cause deployment failures.
Data Flow Dependency Extraction Prompt
For data pipeline architects. Extracts field-level data dependencies from a pipeline description, identifying which upstream fields are consumed by each downstream transformation. Used to prevent broken pipelines from schema changes.
Error Propagation Boundary Prompt Template
For fault-tolerant agent designers. Defines where error propagation should stop in a dependency chain, specifying which downstream tasks should abort, which should run with degraded inputs, and which are fully independent of the failure.
Checkpoint Dependency Graph Prompt
For long-running agent workflows with resume capability. Maps which checkpoints must be preserved and in what order they can be restored, ensuring that resuming from any checkpoint leaves all downstream dependencies satisfiable.
Human-in-the-Loop Handoff Sequencing Prompt
For supervised autonomy workflows. Determines where human approval gates must be inserted into a dependency chain, what information the human needs at each gate, and which downstream tasks are blocked pending approval.
Build Order Generation Prompt for Monorepos
For build system engineers. Generates a correct build order for packages in a monorepo based on import graphs, test dependencies, and publish requirements. Includes detection of missing or circular package dependencies.
Goal Clarification and Constraint Elicitation
Prompt playbooks for refining ambiguous user objectives into well-scoped goals with explicit constraints, success criteria, and boundaries. Useful for agent frontends and planning pre-processors because under-specified goals are the most common root cause of agent loops, over-execution, and misaligned plans.
Ambiguous Goal Refinement Prompt Template
For agent frontends that receive vague user requests. Produces a clarified goal statement with explicit scope, constraints, and success criteria. Includes eval checks for specificity improvement and detection of unresolved ambiguities.
Constraint Elicitation Interview Prompt Template
For planning pre-processors that need to extract implicit constraints from users. Generates a structured interview that probes for security, latency, cost, data, and operational boundaries. Includes completeness scoring and missing-constraint detection.
Success Criteria Definition Prompt Template
For teams building goal-to-plan translators. Converts a high-level objective into measurable, testable success criteria with acceptance thresholds. Includes eval checks for verifiability and alignment with the stated goal.
Scope Boundary Negotiation Prompt Template
For agent orchestrators that must define what is explicitly out of scope. Produces a boundary document with in-scope, out-of-scope, and conditional items. Includes checks for boundary creep and ambiguous edge cases.
User Intent Disambiguation Prompt Template
For conversational agents that encounter ambiguous or conflicting user instructions. Generates targeted clarification questions ranked by impact on execution. Includes eval checks for question necessity and resolution effectiveness.
Goal-to-Requirements Translation Prompt Template
For planning modules that convert clarified goals into structured requirement lists. Produces functional and non-functional requirements with traceability back to the original goal. Includes completeness and consistency validation.
Constraint Conflict Resolution Prompt Template
For agent planners that detect mutually exclusive constraints. Produces a conflict report with resolution options, trade-off analysis, and a recommended constraint hierarchy. Includes checks for hidden conflicts and resolution feasibility.
Implicit Assumption Extraction Prompt Template
For planning pre-processors that surface unstated assumptions before execution begins. Produces a catalog of assumptions with risk ratings and verification steps. Includes eval checks for assumption coverage and criticality assessment.
Risk Boundary Elicitation Prompt Template
For agent operators who need to define acceptable risk thresholds before autonomous execution. Produces a risk boundary specification covering data exposure, action authority, cost limits, and failure tolerance. Includes boundary-gap detection.
Acceptance Criteria Generation Prompt Template
For teams translating requirements into testable acceptance criteria. Produces criteria with preconditions, test steps, and expected outcomes. Includes eval checks for testability and requirement coverage.
Goal Feasibility Assessment Prompt Template
For planning modules that must evaluate whether a goal is achievable given available tools, constraints, and capabilities. Produces a feasibility report with blockers, gaps, and confidence estimates. Includes checks for over-optimistic assessments.
Constraint Completeness Check Prompt Template
For agent pre-processors that validate whether all necessary constraint categories are covered before planning. Produces a gap report organized by constraint domain. Includes eval checks for missing security, privacy, operational, and output constraints.
Success Metric Operationalization Prompt Template
For teams that need to convert qualitative success criteria into measurable metrics. Produces metric definitions with collection method, threshold values, and measurement cadence. Includes checks for metric feasibility and proxy quality.
Edge Case Elicitation Prompt Template
For planning modules that must surface boundary conditions before execution. Produces a structured edge case catalog with expected behavior and handling strategy. Includes coverage checks against the scope boundary.
Domain Constraint Extraction Prompt Template
For agents operating in regulated or specialized domains. Extracts domain-specific constraints from reference material and applies them to the current goal. Includes checks for constraint applicability and regulatory coverage gaps.
Human-in-the-Loop Constraint Definition Prompt Template
For agent architects defining when and how human approval is required. Produces an approval threshold specification with trigger conditions, escalation paths, and timeout behavior. Includes checks for over-delegation and approval-loop bottlenecks.
Hallucination Tolerance Constraint Prompt Template
For teams that need to specify acceptable factual accuracy boundaries for different task types. Produces a tolerance specification with per-category thresholds and verification requirements. Includes checks for misaligned tolerance levels.
Confidence Threshold Definition Prompt Template
For agent runtimes that need explicit confidence boundaries for autonomous action. Produces threshold rules mapping confidence levels to allowed actions, required confirmations, and escalation triggers. Includes calibration checks.
Fallback Behavior Constraint Prompt Template
For resilient agent systems that must define graceful degradation paths. Produces a fallback specification with trigger conditions, degraded-mode behaviors, and recovery criteria. Includes checks for silent failure modes.
Graceful Degradation Boundary Prompt Template
For production agent deployments that need explicit degradation states. Produces a degradation ladder with per-state capabilities, user communication templates, and restoration conditions. Includes checks for unacceptable degradation levels.
Tool Availability Constraint Prompt Template
For agent planners that must adapt goals to available tool capabilities. Produces a tool-to-requirement mapping with gap identification and workaround proposals. Includes checks for tool assumption errors.
Model Capability Boundary Prompt Template
For teams that need to align goals with actual model capabilities. Produces a capability assessment mapping goal requirements to known model strengths and limitations. Includes checks for capability overestimation.
Data Sensitivity Constraint Prompt Template
For agents handling potentially sensitive data. Produces a data classification and handling specification with per-category access rules, retention limits, and redaction requirements. Includes checks for classification gaps.
Output Format Constraint Prompt Template
For planning modules that must enforce output structure requirements. Produces a format specification with schema, validation rules, and compliance checks. Includes checks for format ambiguity and downstream compatibility.
Uncertainty Communication Requirement Prompt Template
For agents that must communicate confidence and uncertainty to users or downstream systems. Produces an uncertainty communication protocol with confidence language, evidence flags, and disclaimer rules. Includes checks for misleading certainty.
Plan Generation and Template Instantiation
Prompt playbooks for producing a complete multi-step plan from a clarified goal, available tools, and known constraints. Useful for planning modules in agent frameworks because the plan is the contract that execution loops, monitoring, and replanning all depend on.
Goal-to-Plan Generation Prompt Template
For agent orchestrators that need a complete, executable multi-step plan from a user goal. Produces a structured plan with step definitions, tool assignments, and success criteria. Includes eval checks for plan completeness, tool coverage gaps, and missing preconditions.
Tool-Aware Plan Generation System Prompt
For planning modules that must generate plans using only available tools and their capabilities. Produces a plan where every step maps to a specific tool or tool combination. Includes harness checks for hallucinated tools, capability mismatches, and argument schema violations.
Constraint-Elicitation Prompt for Agent Goals
For agent frontends that need to extract implicit constraints from ambiguous user requests before planning. Produces a clarified goal with explicit boundaries, budgets, deadlines, and permission scopes. Includes eval criteria for constraint completeness and ambiguity reduction.
Plan Generation with Precondition Checks Prompt
For safety-conscious agent operators who need plans that verify inputs, permissions, and tool availability before each step. Produces a plan with embedded precondition validation gates. Includes failure mode tests for skipped checks and missing dependency detection.
Plan Generation with Postcondition Contracts Prompt
For agents operating on mutable resources where partial failures must be caught. Produces a plan where each step defines expected outputs, side effects, and state changes. Includes validation harness for detecting silent failures and state corruption across steps.
Plan Generation with Human Approval Gates Prompt
For semi-autonomous agent workflows that require human review before high-risk or irreversible actions. Produces a plan with explicit approval checkpoints, review criteria, and escalation paths. Includes eval checks for missing gates on dangerous operations.
Plan Generation with Error Handling Strategy Prompt
For production agent deployments that need retry, skip, escalate, or abort logic per step. Produces a plan with embedded error handling policies, retry budgets, and degradation paths. Includes failure mode tests for infinite retry loops and silent stalls.
Plan Generation with Dependency Graph Output Prompt
For workflow engine developers who need a machine-readable execution graph. Produces a plan with explicit prerequisite relationships, parallel execution groups, and critical path identification. Includes eval checks for cycles, missing edges, and deadlock conditions.
Plan Generation with Checkpoint Insertion Prompt
For long-running autonomous workflows that must survive interruptions, model switches, or context-window overflow. Produces a plan with strategic state-save points and resume instructions. Includes harness tests for state loss scenarios and partial replay costs.
Plan Generation with Uncertainty Markers Prompt
For agents operating with incomplete information where unstated assumptions cause confident but wrong execution. Produces a plan with explicit confidence scores, assumption logging, and clarification triggers for low-certainty steps. Includes eval criteria for assumption detection and appropriate hedging.
Plan Generation with Budget Constraints Prompt
For cost-aware agent deployments that must stay within token, latency, or monetary budgets. Produces a plan with per-step cost estimates, budget allocation, and early termination conditions. Includes harness checks for budget overruns and missing cost tracking.
Plan Generation with Rollback and Cleanup Steps Prompt
For agents that modify persistent state and need safe reversal paths when plans fail mid-execution. Produces a plan with compensating actions, cleanup procedures, and idempotency guarantees. Includes failure mode tests for orphaned resources and partial state corruption.
Plan Generation with Progress Reporting Hooks Prompt
For supervised autonomy where operators need visibility without polling full traces. Produces a plan with embedded status update triggers, milestone definitions, and notification formats. Includes eval checks for missing progress signals and stale status reporting.
Plan Generation with Output Schema Contracts Prompt
For planning modules that must produce plans consumable by downstream execution engines. Produces a plan in a strict, validated schema with typed fields for steps, tools, arguments, and conditions. Includes schema validation harness and field-level completeness checks.
Plan Generation with Timeout Constraints Prompt
For latency-sensitive agent workflows where step duration must be bounded. Produces a plan with per-step and aggregate timeout budgets, deadline propagation, and timeout escalation logic. Includes eval tests for unbounded waits and missing deadline handling.
Plan Generation with Rate Limit Awareness Prompt
For agents calling rate-limited APIs where naive execution triggers throttling or bans. Produces a plan with rate limit budgets, backoff strategies, and request scheduling. Includes harness checks for burst patterns and missing retry-after handling.
Plan Generation with Fallback Tool Specification Prompt
For resilient agent deployments where primary tools may be unavailable. Produces a plan with explicit fallback tool chains, capability degradation paths, and minimum viable tool requirements. Includes failure mode tests for missing fallbacks and cascading unavailability.
Plan Generation with Validation Step Insertion Prompt
For agents where output quality must be verified before proceeding to downstream steps. Produces a plan with embedded validation gates, quality thresholds, and conditional branching on validation results. Includes eval criteria for missing validation on critical outputs.
Plan Generation with Parallel Execution Hints Prompt
For workflow engines that can execute independent steps concurrently to reduce latency. Produces a plan with explicit parallel groups, synchronization barriers, and resource conflict warnings. Includes harness checks for false dependencies and unsafe parallelism.
Plan Generation with Escalation Paths Prompt
For agent deployments where certain failure modes require human intervention or higher-authority decisions. Produces a plan with defined escalation triggers, routing targets, and context preservation for handoff. Includes eval tests for missing escalation on critical failures.
Dynamic Replanning and Plan Revision
Prompt playbooks for updating an in-progress plan when new evidence, tool failures, or unexpected outputs invalidate the original assumptions. Useful for resilient agent runtimes because static plans fail silently in production; replanning prompts define when and how to safely change course.
Agent Replanning Decision Prompt Template
For agent runtime developers building resilient execution loops. Produces a structured decision—continue, patch, rollback, or abort—when a step fails or new evidence contradicts the plan. Includes eval checks for decision consistency, cost-awareness, and safe default behavior.
Agent Plan Patching Prompt for Tool Failure
For agent operators handling runtime tool errors. Generates a minimal plan patch that replaces a failed tool call with an alternative tool, adjusted arguments, or a workaround sequence. Includes harness checks for tool availability, schema compatibility, and idempotency risk.
Agent Plan Patching Prompt for Unexpected Output
For agent runtimes that must recover when a step produces output outside the expected schema or semantic range. Produces a patch that adds validation, reinterprets the output, or branches the plan. Includes eval criteria for output contract adherence and downstream step compatibility.
Agent Plan Patching Prompt for New Evidence
For agents operating with incomplete information where mid-plan discoveries invalidate prior assumptions. Generates a plan revision that incorporates the new evidence, retracts contradicted steps, and re-sequences remaining work. Includes tests for evidence grounding and assumption traceability.
Agent Plan Patching Prompt for Constraint Violation
For safety-conscious agent deployments where budget, time, permission, or policy limits are breached mid-execution. Produces a constraint-aware patch that brings the plan back within bounds or escalates for human approval. Includes eval checks for constraint satisfaction and violation severity classification.
Agent Plan Patching Prompt for Human Override
For human-in-the-loop systems where an operator rejects a step, modifies a goal, or redirects execution. Generates a plan patch that respects the override while preserving completed work and minimizing rework. Includes harness checks for override scope, downstream impact analysis, and audit trail preservation.
Agent Plan Patching Prompt for Partial Completion
For long-running agent workflows where a step partially succeeds—some side effects applied, others failed. Produces a recovery plan that identifies what completed, what must be retried or compensated, and whether rollback is required. Includes eval criteria for state accuracy and compensation completeness.
Agent Plan Patching Prompt for Deadlock Detection
For multi-agent or concurrent execution systems where tasks block each other. Generates a patch that breaks the deadlock through reordering, resource release, or escalation. Includes harness checks for dependency graph correctness and liveness property verification.
Agent Plan Patching Prompt for Context Window Exhaustion
For agents hitting token limits mid-plan where context must be compressed without losing critical state. Produces a summarization and plan continuation strategy that preserves active goals, completed steps, and pending dependencies. Includes eval checks for information preservation and plan coherence after compression.
Agent Plan Patching Prompt for Model Hallucination Detection
For agent runtimes that detect a prior step produced factually inconsistent or fabricated output. Generates a patch that marks the step as unreliable, requests re-execution with grounding, or falls back to verified sources. Includes tests for hallucination classification and correction completeness.
Agent Plan Patching Prompt for Safety Policy Trigger
For regulated or safety-critical agents where a step triggers a content policy, ethical boundary, or compliance rule mid-execution. Produces a patch that halts the violating branch, logs the trigger, and either reroutes or escalates. Includes eval criteria for policy alignment and appropriate refusal behavior.
Agent Plan Patching Prompt for Sub-Agent Failure
For orchestrator agents that delegate to specialized sub-agents and must recover when a sub-agent returns an error, timeout, or incomplete result. Generates a patch that retries, reassigns, or absorbs the sub-task into the parent plan. Includes harness checks for handoff contract validation and error propagation control.
Agent Plan Patching Prompt for Goal Reprioritization
For agents operating under shifting business priorities where the objective function changes mid-execution. Produces a plan revision that reweights remaining tasks, drops deprioritized work, and reallocates resources without corrupting in-progress steps. Includes eval checks for priority consistency and wasted-work minimization.
Agent Plan Patching Prompt for Assumption Invalidation
For agents that explicitly track planning assumptions and must replan when a core assumption is contradicted by evidence. Generates a patch that identifies all assumption-dependent steps, marks them for re-evaluation, and rebuilds the affected plan segment. Includes tests for assumption traceability and impact radius calculation.
Agent Plan Patching Prompt for Loop Detection
For agent runtimes that detect repetitive step patterns indicating a planning or execution loop. Produces a patch that breaks the cycle through step reordering, constraint addition, or escalation. Includes harness checks for loop classification, root cause identification, and termination guarantee.
Agent Plan Patching Prompt for Rollback Requirement
For agents operating on mutable resources where a mid-plan failure requires undoing prior side effects. Generates a compensation plan that identifies reversible steps, orders rollback operations, and preserves audit evidence. Includes eval criteria for rollback completeness and state consistency verification.
Agent Plan Patching Prompt for Authentication Expiry
For long-running agents that lose credential validity mid-execution. Produces a patch that pauses the plan, triggers re-authentication, and resumes or checkpoints based on session recovery success. Includes harness checks for credential scope validation and secure state preservation during auth gaps.
Agent Plan Patching Prompt for Rate Limit
For agents hitting API rate limits that must adapt execution pacing without failing the overall plan. Generates a patch that inserts backoff, batches requests, or switches to a lower-throughput alternative. Includes eval checks for deadline feasibility and throughput estimation accuracy.
Agent Plan Patching Prompt for Validation Failure
For agents where a postcondition check reveals a completed step did not meet quality or correctness criteria. Produces a patch that diagnoses the failure mode, decides between retry with adjusted parameters or alternative approach, and updates downstream step expectations. Includes tests for failure classification and correction strategy appropriateness.
Agent Plan Patching Prompt for Checkpoint Restoration Failure
For agents that attempt to resume from a saved checkpoint but find the state corrupted, incomplete, or incompatible. Generates a recovery plan that identifies the last known-good state, estimates lost progress, and rebuilds the execution path. Includes harness checks for state integrity validation and minimal re-execution.
Progress Tracking and State Summarization
Prompt playbooks for maintaining an accurate representation of what has been completed, what is in progress, and what remains. Useful for long-running agent sessions and multi-turn workflows because stale or incomplete state summaries cause repeated work, dropped subtasks, and context-window pollution.
Agent Task Progress Tracker Prompt Template
For agent orchestrators that need a structured snapshot of completed, in-progress, and pending subtasks. Produces a machine-readable progress record with status, timestamps, and blocker flags. Includes eval checks for completeness and stale-state detection.
Multi-Turn Session State Summary Prompt Template
For long-running conversational agents that must compress conversation history into a dense state summary. Produces a structured summary capturing resolved items, open questions, user preferences, and pending actions. Includes fidelity checks against the full transcript.
Checkpoint State Capture Prompt for Agent Sessions
For agents that need to serialize their full state at deliberate save points so execution can resume after interruption or model switch. Produces a complete, restorable state object. Includes validation for state integrity and missing-field detection.
Context Window State Compression Prompt Template
For agents approaching token limits that must decide what to retain, summarize, or evict from the context window. Produces a pruned context with salience annotations and eviction rationale. Includes eval checks for critical information loss.
Agent Memory Refresh and State Reconciliation Prompt
For agents returning to a session after a gap and needing to reconcile stored state with new observations. Produces an updated state with conflict flags and stale-data markers. Includes tests for detecting contradictions between memory and fresh evidence.
In-Progress Task Blocker Detection Prompt Template
For agent runtimes that need to identify why a task is stalled and classify the blocker type. Produces a blocker report with root cause, affected subtasks, and suggested resolution path. Includes eval criteria for false-positive blocker flags.
Agent Action Log to Structured Status Prompt
For converting raw agent action logs and tool-call histories into a structured progress summary. Produces a timeline of completed actions with outcomes, errors, and state changes. Includes validation for log-to-summary consistency.
Stale State Detection and Correction Prompt Template
For agents that must detect when their internal state no longer matches ground truth and propose corrections. Produces a diff between believed state and observed state with correction actions. Includes tests for over-correction and confirmation bias.
Session Handoff State Summary Prompt Template
For multi-agent or human-handoff scenarios where one agent must transfer its full understanding to another. Produces a handoff package with current goal, completed work, open items, assumptions, and risk flags. Includes completeness checks for the receiving agent.
Error and Retry State Documentation Prompt
For agents that encounter failures and must document what was attempted, what failed, and what state was preserved before retry or escalation. Produces an error-state record suitable for debugging and retry logic. Includes eval checks for missing failure context.
Agent State Diff Generation Prompt Template
For comparing two agent state snapshots and producing a structured diff of what changed. Produces a change log with added, removed, and modified fields plus confidence scores. Includes tests for detecting silent state corruption.
Context Pollution Detection and Cleanup Prompt
For agents whose context window has accumulated irrelevant, contradictory, or outdated information that degrades reasoning. Produces a contamination report with cleanup recommendations. Includes eval checks for false-positive pollution flags.
Multi-Agent Shared State Consolidation Prompt
For systems where multiple agents contribute to a shared task state and need a single coherent view. Produces a merged state resolving conflicts, duplicates, and gaps across agent contributions. Includes consistency validation across source agents.
State Serialization for Agent Restart Prompt Template
For agents that must persist their full state to durable storage so a new instance can resume identically. Produces a serialized state blob with version marker and integrity hash. Includes deserialization round-trip validation.
Agent Workflow Audit Trail Prompt Template
For generating a human-auditable record of agent decisions, state transitions, and tool calls across a workflow. Produces a timestamped audit log with decision rationale and evidence links. Includes completeness checks against the raw execution trace.
State Reconciliation After Tool Failure Prompt
For agents that must repair their state after a tool call returns an error, partial result, or timeout. Produces a reconciled state with uncertainty markers for unverified assumptions. Includes tests for cascading state corruption.
Agent Loop Detection and State Reset Prompt
For detecting when an agent is repeating actions without progress and proposing a state reset or strategy change. Produces a loop diagnosis with evidence, affected state fields, and reset recommendations. Includes false-positive loop detection checks.
State Integrity Validation Prompt Template
For verifying that an agent's internal state is internally consistent, complete, and free of contradictory entries. Produces a validation report with integrity violations and severity ratings. Includes schema conformance and cross-field consistency checks.
Progress Blockage Escalation Prompt Template
For agents that have exhausted autonomous recovery options and must escalate a blocked task to a human operator. Produces an escalation summary with what was attempted, why it failed, current state, and what decision is needed. Includes clarity and actionability checks.
State Summarization with Confidence Scoring Prompt
For agents that must annotate each element of their state summary with a confidence level, especially when operating on inferred or uncertain information. Produces a state summary with per-field confidence scores and evidence provenance. Includes calibration checks against ground truth.
Agent State Recovery from Incomplete Log Prompt
For reconstructing agent state when the execution log has gaps, missing entries, or corrupted records. Produces a best-effort reconstructed state with explicit gap markers and assumptions. Includes tests for over-confident reconstruction.
Progress Tracking with Uncertainty Quantification Prompt
For agents that must report progress as a range or distribution rather than a false-precision percentage. Produces a progress estimate with uncertainty bounds, confidence intervals, and key unknowns. Includes eval checks for overconfident completion estimates.
State-Based Next Action Recommendation Prompt
For agents that must analyze their current state and recommend the highest-priority next action. Produces a ranked list of candidate actions with rationale, expected state change, and risk flags. Includes tests for action selection against ground-truth priorities.
Precondition Validation and Readiness Checks
Prompt playbooks for verifying that required inputs, permissions, tool availability, and environmental conditions are met before executing a step. Useful for safety-conscious agent operators because skipping precondition checks leads to runtime errors that are harder to recover from than upfront refusal.
Tool Availability Probe Prompt Template
For agent operators who need to verify that required tools, APIs, and MCP servers are reachable before starting a task. Produces a readiness report with per-tool status, capability gaps, and a go/no-go recommendation. Includes eval checks for false-positive availability claims and missing dependency detection.
API Key and Authentication Status Check Prompt
For platform engineers building agent harnesses that must validate credentials before execution. Produces an auth status summary across required services with expiry warnings and scope verification. Includes harness concerns for secret redaction in logs and token freshness checks.
Input Data Schema Validation Prompt Template
For data pipeline teams who need to confirm that incoming payloads match expected schemas before agent processing. Produces a field-by-field conformance report with type mismatches, missing required fields, and unexpected keys flagged. Includes eval criteria for schema drift detection and null handling.
Required Field Presence Check Prompt
For integration engineers validating that all mandatory inputs exist before an agent step executes. Produces a completeness checklist with explicit pass/fail per field and remediation suggestions for missing data. Includes harness logic for early bail-out and structured error responses.
Context Window Capacity Check Prompt
For agent runtime developers who need to verify that the assembled prompt, tools, and conversation history fit within the model's context limit. Produces a token budget breakdown with headroom analysis and truncation warnings. Includes eval checks for silent context overflow and priority-based trimming recommendations.
Rate Limit Budget Check Prompt
For production operators managing API rate limits across multiple agent steps. Produces a remaining-quota assessment with per-endpoint budgets, estimated step consumption, and risk of mid-plan exhaustion. Includes harness concerns for dynamic backoff and quota refresh timing.
Model Endpoint Health Check Prompt
For SRE teams monitoring inference endpoint availability before dispatching agent workloads. Produces a health status report covering latency, error rates, and model version consistency. Includes eval checks for stale health data and failover readiness assessment.
Database Connection Readiness Prompt
For data engineers whose agents need to verify database reachability, permissions, and schema compatibility before executing read or write operations. Produces a connection diagnostic with table access audit and credential validity check. Includes harness logic for connection pooling state and timeout handling.
File System Permission Verification Prompt
For infrastructure teams running agents that read or write to local or mounted file systems. Produces a permission matrix per path with read, write, and execute status for the agent's effective user. Includes eval checks for symlink traversal risks and mount point availability.
Environment Variable Validation Prompt
For DevOps engineers ensuring that all required environment variables are set and well-formed before agent initialization. Produces a variable audit with type checks, format validation, and sensitive-value presence confirmation without exposing secrets. Includes harness concerns for default fallback detection.
Dependency Vulnerability Audit Prompt
For security-conscious teams who need to check that agent dependencies have no known critical vulnerabilities before execution in sensitive environments. Produces a risk summary with CVE references, severity ratings, and remediation urgency. Includes eval checks for false negatives from stale advisory databases.
Certificate Expiry and Validity Check Prompt
For platform teams whose agents connect to TLS-secured endpoints and must avoid mid-execution certificate failures. Produces a certificate health report with expiry dates, chain validation, and hostname mismatch detection. Includes harness logic for short-lived certificate renewal windows.
Secrets Manager Accessibility Check Prompt
For security engineers verifying that the agent runtime can retrieve required secrets from vault services before attempting authenticated operations. Produces an access report per secret path with permission status and rotation schedule awareness. Includes harness concerns for secret caching staleness.
External Service Dependency Health Prompt
For platform operators who need a unified readiness view across all third-party services an agent plan depends on. Produces a dependency health matrix with per-service status, SLA compliance, and degraded-mode recommendations. Includes eval checks for cascading failure risk and partial-degradation handling.
Feature Flag Rollout Status Prompt
For release engineers using feature flags to gate agent capabilities in production. Produces a flag evaluation report showing which agent behaviors are enabled, disabled, or in gradual rollout for the current context. Includes harness concerns for flag evaluation consistency across retries.
User Identity and Session Validity Prompt
For auth platform teams who need to confirm that the requesting user's session is valid, non-expired, and has not been revoked before the agent acts on their behalf. Produces a session diagnostic with scope verification and step-up auth requirements. Includes eval checks for token replay detection.
Data Classification and Sensitivity Check Prompt
For data governance teams who need to verify the classification level of input data before the agent processes it, ensuring handling rules match policy. Produces a classification label with handling constraints, residency requirements, and allowed processing boundaries. Includes harness logic for automatic redaction triggers.
Compliance Policy Adherence Check Prompt
For compliance officers embedding policy gates into agent workflows. Produces a policy conformance report mapping the planned action against applicable rules, with violations flagged and required approvals listed. Includes eval checks for policy version staleness and regulatory change detection.
Circuit Breaker State Verification Prompt
For resilience engineers who need to check whether circuit breakers protecting downstream services are open before the agent attempts a call. Produces a breaker state summary with failure counts, cooldown status, and fallback availability. Includes harness logic for half-open probing and graceful degradation paths.
Idempotency Key Validation Prompt
For payment and transactional system engineers who must ensure idempotency keys are present and well-formed before the agent executes non-idempotent operations. Produces a key validity report with collision risk assessment and replay detection window status. Includes harness concerns for key generation entropy checks.
Human Approval Queue Status Prompt
For human-in-the-loop system designers who need to verify that approval queues are available and staffed before the agent submits a review request. Produces a queue health report with estimated wait time, on-call coverage, and escalation path readiness. Includes eval checks for SLA breach risk on pending approvals.
Prompt Injection Defense Readiness Check Prompt
For security engineers validating that input sanitization and instruction hierarchy defenses are active before processing untrusted content. Produces a defense posture report with active guard layers, detected injection patterns, and residual risk assessment. Includes harness logic for content quarantine and safe preview modes.
Safety Policy Boundary Clarification Prompt
For trust and safety teams who need to pre-evaluate whether a planned agent action falls within allowed policy boundaries before execution. Produces a boundary assessment with policy citations, grey-area flags, and recommended escalation path. Includes eval checks for policy interpretation consistency across similar requests.
Postcondition Validation and Step Verification
Prompt playbooks for confirming that a completed step produced the expected output, side effects, and state changes before proceeding. Useful for agents operating on mutable resources because undetected partial failures compound across steps and corrupt downstream reasoning.
Step Completion Confirmation Prompt Template
For agent orchestrators that need to verify a step finished before proceeding. Produces a structured confirmation with success/failure status, evidence summary, and next-step gating decision. Includes eval checks for false-positive confirmations when tools return ambiguous signals.
Output Contract Validation Prompt for Agent Steps
For agent developers who define expected output schemas per step. Compares actual step output against a declared contract and produces a pass/fail verdict with field-level mismatch details. Includes harness concerns for schema drift and partial match scoring.
Side Effect Detection Prompt for Mutable Resources
For operators of agents that modify databases, filesystems, or cloud resources. Produces a catalog of detected side effects beyond the primary output, with severity ratings and rollback recommendations. Includes eval checks for missed side effects that corrupt downstream state.
State Change Verification Prompt After Tool Execution
For agent runtimes that need to confirm a tool call actually changed system state. Compares pre- and post-execution state snapshots and produces a change summary with unexpected delta flags. Includes harness concerns for noisy state representations and timing races.
Partial Failure Detection Prompt for Multi-Step Agents
For agent operators running sequential tool chains where intermediate steps can silently fail. Produces a step-by-step health report identifying which substeps succeeded, failed, or produced degraded results. Includes eval checks for cascading failure blindness.
Downstream Corruption Guard Prompt Template
For agent architects who need to prevent one bad step from poisoning subsequent reasoning. Evaluates whether the current step output contains errors, hallucinations, or malformed data that would corrupt the next step's input. Includes harness concerns for over-blocking valid partial results.
Step Rollback Decision Prompt After Validation Failure
For agent runtimes that must decide whether to roll back, retry, skip, or abort after a step fails validation. Produces a structured decision with rationale, rollback scope, and cleanup actions. Includes eval checks for inappropriate rollback scope and resource leak risks.
Precondition vs Postcondition Comparison Prompt
For agent developers who need to verify that a step's actual postconditions match the expected ones defined before execution. Produces a diff between expected and actual postconditions with gap analysis. Includes harness concerns for implicit postconditions the model may miss.
Tool Output Schema Conformance Check Prompt
For teams integrating external tools where output shapes vary. Validates that tool output conforms to the expected JSON schema, type constraints, and required fields before the agent consumes it. Includes eval checks for silent type coercion and missing null handling.
Agent Action Audit Trail Verification Prompt
For compliance and debugging workflows that need to confirm every agent action was logged. Compares the declared action log against expected entries and produces a completeness report with missing or tampered records. Includes harness concerns for log format inconsistency.
Idempotency Check Prompt for Repeated Steps
For agent operators who need to confirm that re-executing a step is safe. Analyzes whether a step has already been applied and produces a safe-to-repeat verdict with evidence of prior completion. Includes eval checks for false idempotency claims on non-idempotent operations.
Data Integrity Check Prompt After Agent Mutation
For agents that write to databases or data stores and must verify consistency afterward. Produces an integrity report checking referential integrity, constraint violations, and unexpected nulls after the agent's write. Includes harness concerns for partial writes and transaction boundaries.
Expected Side Effect Enumeration Prompt
For agent planners that need to declare anticipated side effects before step execution so post-hoc verification has a baseline. Produces a structured list of expected side effects with resource targets and acceptable variance. Includes eval checks for over-narrow or over-broad enumeration.
Unexpected Side Effect Discovery Prompt
For post-execution analysis when agents modify shared environments. Compares actual system state changes against a declared expected-side-effect list and flags anomalies. Includes harness concerns for detection sensitivity and false-positive noise in dynamic environments.
Validation Harness Prompt for Agent Pipelines
For teams building reusable validation layers across agent steps. Produces a composable validation plan that chains multiple postcondition checks into a single pass/fail gate with aggregated evidence. Includes eval concerns for harness composition errors and check ordering dependencies.
Output Shape Assertion Prompt for JSON Steps
For agent developers who need strict structural validation on step outputs. Produces a structural assertion report checking key presence, nesting depth, array lengths, and type consistency without evaluating semantic correctness. Includes harness concerns for valid-but-wrong content passing structural checks.
Null and Empty Value Detection Prompt
For agents operating on data pipelines where missing values silently propagate. Produces a field-level null/empty audit of step output with severity classification and imputation recommendations. Includes eval checks for semantically meaningful empty values being incorrectly flagged.
Record Existence Confirmation Prompt After Insert
For agents that create database records and must verify persistence before proceeding. Produces a confirmation with record identity, retrieval evidence, and field comparison against the insert payload. Includes harness concerns for eventual consistency delays and replication lag.
Record Update Field Comparison Prompt
For agents that modify existing records and need to confirm only intended fields changed. Produces a before/after field diff with unintended change flags and rollback suggestions. Includes eval checks for timestamp and audit-field noise masking real changes.
Deployment Health Verification Prompt After Agent Action
For DevOps agents that trigger deployments and must confirm service health before marking the step complete. Produces a health assessment from metrics, logs, and endpoint checks with a go/no-go recommendation. Includes harness concerns for warm-up periods and transient errors.
End-to-End Workflow Success Assertion Prompt
For agent orchestrators that need a final gate confirming the entire multi-step workflow achieved its objective. Produces a summary assertion comparing the final state against the original goal with per-step contribution analysis. Includes eval checks for premature success declarations when side effects are unresolved.
Error Handling and Fallback Strategy
Prompt playbooks for deciding what an agent should do when a step fails, including retry, skip, escalate, or abort logic. Useful for production agent deployments because unhandled errors cause silent stalls or infinite retry loops that waste tokens and violate latency budgets.
Agent Step Failure Classification Prompt Template
For agent runtime developers who need to classify step failures into actionable categories (transient, permanent, permission, timeout, malformed output). Produces a structured failure record with retry eligibility, severity, and recommended recovery action. Includes eval checks for classification accuracy against known failure modes.
Retry Decision Prompt for Failed Tool Calls
For agent orchestrators deciding whether to retry a failed tool call. Evaluates error type, retry count, token budget remaining, and step criticality to output a retry/abort/skip decision with backoff timing. Includes harness checks for infinite loop prevention and budget exhaustion.
Skip vs Abort Decision Prompt for Agent Workflows
For workflow engines that must decide whether a failed step is optional enough to skip or critical enough to abort the entire plan. Produces a decision with dependency impact analysis and downstream risk assessment. Evaluation criteria include false-skip detection and cascade failure prevention.
Escalation Trigger Prompt for Autonomous Agents
For supervised autonomy systems that need to determine when a failure pattern requires human intervention. Outputs an escalation decision with severity justification, context summary, and recommended human action. Includes thresholds for confidence, retry exhaustion, and safety impact.
Error Severity Scoring Prompt for Agent Steps
For monitoring systems that need consistent severity scores across heterogeneous agent failures. Produces a severity level (critical/high/medium/low) with blast radius estimate, data loss risk, and downstream step impact. Eval checks for inter-rater consistency and severity inflation.
Fallback Tool Selection Prompt for Unavailable APIs
For agent tool-use layers when a primary tool or API is unavailable. Selects the best alternative tool from a capability registry, maps parameters, and flags capability gaps. Includes validation that fallback tools satisfy minimum requirements and don't silently degrade output quality.
Graceful Degradation Prompt for Partial Agent Failures
For user-facing agents that must produce useful output even when some steps fail. Generates a partial result with explicit caveats, missing sections flagged, and confidence annotations. Evaluation checks that degradation is transparent and doesn't fabricate missing data.
Timeout Recovery Prompt for Long-Running Agent Steps
For agent runtimes handling steps that exceed latency budgets. Decides whether to accept partial results, retry with reduced scope, or escalate based on what was completed before timeout. Includes harness for timeout simulation and partial-output validation.
Infinite Loop Detection Prompt for Agent Execution
For agent monitors that detect repetitive tool calls, state cycles, or non-progressing loops. Analyzes recent execution history to classify loop type and output a break action with justification. Eval criteria include false-positive rate on legitimate iterative patterns.
Token Budget Exhaustion Fallback Prompt
For context-window management when an agent approaches token limits mid-execution. Produces a compressed state summary, prioritized remaining steps, and a decision to continue with summarization or abort with partial results. Includes budget-tracking harness.
Agent Step Rollback Prompt Template
For agents operating on mutable resources that need to undo partial side effects after a failure. Generates a rollback plan identifying what changed, reversal steps, and verification checks. Evaluation includes completeness of side-effect detection and safe reversal ordering.
Partial Completion Summary Prompt for Failed Workflows
For generating user-facing summaries when an agent workflow terminates before completion. Produces a structured summary of what was accomplished, what remains, known blockers, and recommended next steps. Eval checks for honesty about incomplete work and actionable handoff.
Error Context Collection Prompt for Debugging Agents
For developer tooling that captures structured debugging context from agent failures. Extracts the failing step, inputs, outputs, error messages, surrounding state, and relevant conversation turns into a debug bundle. Includes completeness checks for root-cause analysis sufficiency.
Human Handoff Prompt for Unresolvable Agent Errors
For agents that must transfer control to a human operator after exhausting automated recovery. Generates a handoff message with failure summary, attempted recoveries, current state, and specific human action requested. Evaluation criteria include clarity, completeness, and appropriate urgency signaling.
Confidence Threshold Fallback Prompt for Low-Certainty Steps
For agents that detect low-confidence outputs before committing to downstream actions. Decides whether to request clarification, use a safer alternative, or escalate based on confidence score and action risk. Includes calibration checks against known-correct outcomes.
Alternative Path Discovery Prompt for Blocked Agent Tasks
For planning agents that must find a different approach when the primary execution path is blocked. Generates alternative subtask sequences using available tools, with feasibility assessment and trade-off comparison. Eval checks that alternatives are genuinely distinct and executable.
Error Pattern Recognition Prompt for Recurring Failures
For agent observability systems analyzing failure logs to identify recurring patterns. Classifies failure clusters, identifies root cause hypotheses, and suggests preventive measures. Includes precision/recall evaluation against known failure categories.
Agent Self-Diagnosis Prompt for Execution Failures
For agents that must explain their own failure to operators or downstream systems. Produces a structured diagnosis including what was attempted, what went wrong, likely causes, and confidence in the diagnosis. Eval checks for honest uncertainty expression and avoidance of fabricated explanations.
Fallback Data Source Prompt for Retrieval Failures
For RAG-capable agents when primary knowledge sources are unavailable or return empty results. Selects alternative sources from a ranked fallback registry, reformulates queries, and flags reduced confidence. Includes validation that fallback sources are appropriate for the domain.
Agent Recovery Plan Generation Prompt
For agent orchestrators that need a multi-step recovery plan after a mid-workflow failure. Produces a sequenced recovery plan with state restoration, retry or alternative steps, and verification gates. Evaluation includes plan feasibility, dependency correctness, and minimal rework.
Retry vs Replan Decision Prompt for Complex Agent Failures
For agents facing failures where simple retry may be insufficient and full replanning may be excessive. Analyzes failure context, plan progress, and remaining goals to decide between retry, local repair, or global replan. Includes cost-benefit evaluation criteria.
Circuit Breaker Activation Prompt for Failing Dependencies
For agent infrastructure that must protect downstream services from cascading retry storms. Decides when to open a circuit breaker based on failure rate, latency degradation, and error types. Outputs breaker state change with cooldown duration and fallback mode. Includes threshold tuning guidance.
Agent Error Recovery SLA Enforcement Prompt
For production agents operating under latency or completion SLAs. Evaluates whether recovery actions can complete within remaining SLA budget and decides between fast-degrade, escalate, or accept SLA violation with documentation. Includes SLA-tracking harness.
Fallback Model Selection Prompt for Primary Model Failure
For model routing layers when the primary model returns errors, exceeds latency, or produces unusable output. Selects a fallback model considering capability match, cost, latency, and output format compatibility. Includes regression checks that fallback quality meets minimum thresholds.
Agent Failure Postmortem Generation Prompt
For generating structured postmortems from agent execution traces after significant failures. Produces timeline, root cause analysis, impact assessment, and preventive recommendations. Evaluation includes factual accuracy against trace data and actionable recommendation quality.
Checkpoint and Context Preservation
Prompt playbooks for saving agent state at deliberate points so execution can resume after interruption, model switch, or context-window overflow. Useful for long-running autonomous workflows and human-in-the-loop systems because losing intermediate state forces expensive full replays.
Agent State Snapshot Prompt Template
For agent framework developers who need to serialize complete agent state at a checkpoint. Produces a structured snapshot including active plan, completed steps, current tool outputs, and pending decisions. Includes eval checks for snapshot completeness and resumability.
Context Window Overflow Recovery Prompt
For operators of long-running agents that hit token limits mid-execution. Produces a compressed state summary with priority-ranked context, dropped detail justification, and a resumption handoff. Includes eval for critical context preservation and information loss detection.
Interruption-Resume Prompt for Multi-Session Agents
For systems where agents must pause and resume across sessions, deployments, or human review cycles. Produces a resumption brief that reconstructs intent, progress, and next actions from saved state. Includes eval checks for intent drift and step omission after resume.
Human-in-the-Loop Handoff State Prompt
For agent workflows that pause at approval gates and hand context to human reviewers. Produces a reviewer-ready summary with pending decisions, action context, risk flags, and expected outcomes. Includes eval for decision clarity and missing consequence disclosure.
Model Switch State Transfer Prompt
For teams migrating agent state between different models or providers mid-execution. Produces a model-agnostic state representation with tool call histories, plan progress, and unresolved dependencies. Includes eval for tool schema compatibility and reasoning fidelity across model boundaries.
Context Pruning and State Preservation Prompt
For agents that must aggressively reduce context while preserving execution continuity. Produces a pruned context with salience-ranked retention, explicit removal rationale, and a recovery path for pruned information. Includes eval for critical fact retention and downstream error introduction.
Resume-from-Checkpoint Validation Prompt
For agent runtimes that must verify checkpoint integrity before resuming execution. Produces a validation report confirming state consistency, tool availability, plan alignment, and environmental prerequisites. Includes eval for false-positive resumption and silent state corruption detection.
Agent Context Reconstruction Prompt After Failure
For operators recovering agent state after crashes, timeouts, or partial execution failures. Produces a reconstructed execution trace from logs, tool outputs, and partial state with confidence markers on inferred versus confirmed state. Includes eval for reconstruction accuracy and missing step identification.
Checkpoint Frequency Decision Prompt
For agent orchestrators that need to decide when to save state during execution. Produces a checkpoint schedule based on step risk, reversibility, cost of replay, and context window pressure. Includes eval for over-saving waste and under-saving replay cost balance.
State Diff Generation Prompt for Incremental Saves
For systems implementing incremental checkpointing to reduce storage and token overhead. Produces a minimal state diff showing only changes since the last checkpoint with base state reference. Includes eval for diff completeness and correct base state anchoring.
Multi-Agent Handoff State Preservation Prompt
For multi-agent systems where one agent passes execution context to another specialized agent. Produces a handoff package with task state, relevant history, unresolved dependencies, and authority boundaries. Includes eval for context sufficiency and duplicate work prevention.
Checkpoint Integrity Verification Prompt
For systems that must cryptographically or logically verify checkpoint data hasn't been corrupted or tampered with. Produces an integrity report with hash verification, schema compliance, and logical consistency checks. Includes eval for false-positive integrity passes and corruption detection sensitivity.
State Rollback Prompt After Invalid Step
For agents that must undo partial execution after detecting an invalid or harmful step. Produces a rollback plan identifying affected state, reversal actions, and a safe resumption point. Includes eval for incomplete rollback detection and side-effect coverage.
Checkpoint Compression for Token Budget Prompt
For agents operating under strict token budgets that must compress checkpoints without losing resumability. Produces a compressed checkpoint with priority-tiered detail, lossy compression rationale, and decompression hints. Includes eval for compression ratio versus resumption accuracy trade-off.
Resume with Updated Instructions Prompt
For agents that must resume execution after system prompts, policies, or goals have changed mid-workflow. Produces a state reconciliation merging prior progress with new instructions, flagging conflicts, and proposing a revised execution path. Includes eval for instruction conflict resolution and stale assumption carryover.
Agent State Summary for Human Review Prompt
For operators and auditors who need a human-readable summary of agent state at any checkpoint. Produces a structured brief covering progress, decisions made, evidence used, risks encountered, and pending actions. Includes eval for completeness, accuracy, and decision traceability.
Checkpoint-Aware Error Recovery Prompt
For agents that must recover from errors by rolling back to a known-good checkpoint and retrying with adjusted strategy. Produces a recovery plan selecting the appropriate checkpoint, diagnosing the failure, and modifying the approach before resumption. Includes eval for correct checkpoint selection and repeated failure prevention.
Graceful Shutdown and State Save Prompt
For agents that must save state and shut down cleanly on signal, timeout, or resource exhaustion. Produces a shutdown checkpoint with in-flight task handling, resource cleanup notes, and priority-ordered resumption queue. Includes eval for in-flight task loss and resumption readiness.
Context Preservation Across Model Upgrades Prompt
For teams upgrading underlying models while agents have active saved state. Produces a compatibility report identifying state elements that may behave differently under the new model and a migration plan with fallback checkpoints. Includes eval for behavioral regression detection and safe upgrade gating.
Checkpoint-Aware Replanning Prompt
For agents that must revise their plan from a checkpoint after discovering new constraints, tool failures, or changed goals. Produces a revised plan that preserves valid completed work, reorders remaining steps, and justifies deviations from the original plan. Includes eval for unnecessary rework detection and plan consistency.
Plan Critique and Self-Evaluation
Prompt playbooks for having the model review its own plan for completeness, feasibility, tool coverage, and risk before execution begins. Useful for planning modules that must catch errors before they become expensive execution failures, especially in high-cost or high-risk agent workflows.
Pre-Execution Plan Completeness Review Prompt Template
For agent orchestrator developers who need to validate a generated plan before execution begins. Produces a structured completeness report identifying missing subtasks, undefined inputs, and unhandled termination conditions. Includes eval criteria for false positives and missed gaps.
Tool Coverage Audit Prompt for Agent Plans
For platform engineers building agent frameworks with bounded tool sets. Audits a proposed plan against available tool manifests to flag steps with no matching capability, insufficient permissions, or missing parameter schemas. Outputs a gap matrix with severity ratings.
Pre-Mortem Failure Mode Enumeration Prompt Template
For reliability engineers deploying autonomous agents in production. Generates a structured list of likely failure modes for a given plan, ranked by impact and probability, before any step executes. Includes harness for tracking whether predicted failures materialize.
Plan Step Dependency Validation Prompt Template
For workflow engine developers who need to verify execution order correctness. Analyzes a plan's step sequence to detect missing prerequisites, circular dependencies, and steps that can be safely parallelized. Outputs a corrected dependency graph with rationale.
Assumption Extraction and Challenge Prompt Template
For agent safety engineers who need to surface hidden assumptions before irreversible actions. Extracts every implicit assumption from a plan, rates each for verifiability, and generates test queries to validate or invalidate them. Includes eval for assumption recall.
Plan vs Goal Alignment Scoring Prompt Template
For planning module evaluators who need to measure whether a generated plan actually satisfies the original objective. Produces a dimension-by-dimension alignment score with specific misalignment citations. Includes rubric calibration guidance.
Edge Case Coverage Review Prompt Template
For QA engineers validating agent plans against boundary conditions. Systematically probes a plan for unhandled edge cases across input ranges, state transitions, and error conditions. Outputs a prioritized list of gaps with reproduction scenarios.
Constraint Satisfaction Check Prompt Template
For agent architects who must ensure plans respect hard constraints like latency budgets, token limits, and compliance rules. Validates each step against a constraint manifest and flags violations with severity and suggested alternatives.
Side Effect Prediction and Review Prompt Template
For operators of agents that mutate production resources. Predicts all observable side effects of a plan including state changes, external service impacts, and data mutations. Outputs a side-effect manifest suitable for pre-approval review gates.
Idempotency Check for Plan Steps Prompt Template
For distributed systems engineers building retry-capable agent runtimes. Reviews each plan step for idempotency properties, identifies steps that would produce duplicate effects on retry, and recommends safe guard patterns. Includes test case generation.
Error Propagation Analysis Prompt Template
For resilience engineers designing agent error handling. Traces how a failure at each plan step would cascade through downstream dependencies, identifying single points of failure and recommending containment boundaries. Outputs a propagation graph.
Fallback Strategy Adequacy Review Prompt Template
For production SRE teams deploying agents with degraded-mode requirements. Evaluates whether the plan's fallback strategies cover all critical failure modes and meet recovery time objectives. Flags gaps where the agent would stall or produce unsafe partial results.
Human Approval Gate Placement Review Prompt Template
For human-in-the-loop system designers who need to decide where approval gates belong. Reviews a plan to identify steps that require human judgment based on risk, irreversibility, and ambiguity criteria. Outputs recommended gate positions with justification.
Context Window Overflow Risk Assessment Prompt Template
For agent runtime developers managing long-running workflows. Estimates token consumption per plan step, identifies points where accumulated context would exceed model limits, and recommends checkpoint or summarization insertion points.
Plan Ambiguity Detection Prompt Template
For planning module developers who need to catch vague steps before they reach execution. Flags plan steps with ambiguous success criteria, underspecified outputs, or unclear tool selection logic. Outputs specific clarification questions for each ambiguity.
Hallucination Vulnerability Audit Prompt for Plans
For AI safety engineers evaluating agent reliability. Identifies plan steps where the model is likely to fabricate tool outputs, invent data, or confabulate reasoning chains. Rates each step's hallucination risk and recommends grounding countermeasures.
Plan Rollback Feasibility Assessment Prompt Template
For infrastructure engineers deploying agents against mutable systems. Evaluates whether each plan step can be safely undone and what rollback procedures would be required. Flags irreversible steps that need explicit confirmation gates.
Cost Estimation and Budget Review Prompt Template
For engineering leads managing agent operating costs. Estimates token consumption, API call counts, and expected latency for a plan before execution. Compares against a budget constraint and flags steps that exceed thresholds with optimization suggestions.
Single Point of Failure Identification Prompt Template
For reliability architects hardening agent workflows. Scans a plan for steps, tools, or dependencies whose failure would block all remaining progress. Outputs a ranked list of SPOFs with recommended redundancy or mitigation strategies.
Plan Security Vulnerability Scan Prompt Template
For security engineers reviewing autonomous agent behavior. Audits a plan for prompt injection surfaces, sensitive data exposure risks, unauthorized tool access paths, and privilege escalation opportunities. Outputs findings with severity and remediation steps.
Plan Monitoring Coverage Review Prompt Template
For observability engineers instrumenting agent runtimes. Reviews a plan to identify steps lacking adequate logging, metrics, or alerting triggers. Outputs a monitoring gap report with specific instrumentation recommendations for each unobserved step.
Plan for Incomplete Data Handling Prompt Template
For data platform engineers building agents that operate on partial or streaming data. Evaluates how each plan step behaves when expected data is missing, stale, or truncated. Flags steps that would silently proceed with degraded inputs and recommends guard logic.
Uncertainty Handling and Assumption Tracking
Prompt playbooks for making planning assumptions explicit, flagging low-confidence steps, and requesting clarification or evidence before committing to irreversible actions. Useful for agents operating with incomplete information because unstated assumptions are the primary source of confident but wrong execution.
Assumption Inventory Generation Prompt Template
For agent planning pre-processors. Extracts all implicit and explicit assumptions from a goal statement and available context. Produces a structured inventory with assumption IDs, categories, and initial confidence estimates. Includes eval checks for completeness and hidden dependency coverage.
Pre-Flight Assumption Check Prompt Template
For agent runtimes before plan execution begins. Validates each planning assumption against current evidence and flags violations. Produces a go/no-go assessment with blocking assumptions and warnings. Includes harness for evidence freshness checks and stale context detection.
Irreversible Action Confirmation Prompt Template
For agents about to execute destructive or non-rollbackable operations. Requires explicit confirmation with enumerated consequences, affected resources, and rollback impossibility. Produces a structured confirmation request with risk summary. Includes eval for false-negative approvals and missing consequence disclosure.
Low-Confidence Step Flagging Prompt Template
For plan review modules before execution. Scores each plan step for confidence based on evidence availability, assumption stability, and tool reliability. Produces a prioritized list of steps requiring additional evidence or human review. Includes calibration checks against actual step outcomes.
Clarification Question Generation Prompt Template
For agent frontends handling ambiguous user instructions. Generates targeted clarification questions ranked by impact on plan correctness. Produces a minimal question set that resolves critical ambiguities without overwhelming the user. Includes eval for question necessity and ambiguity resolution coverage.
Evidence Threshold Decision Prompt Template
For agents deciding whether to proceed, escalate, or gather more evidence. Applies configurable evidence sufficiency thresholds to each decision point. Produces a decision with rationale and specific evidence gaps if threshold not met. Includes harness for threshold configuration and override logging.
Assumption-to-Evidence Mapping Prompt Template
For audit and traceability workflows. Maps each planning assumption to supporting or contradicting evidence sources. Produces a traceability matrix with evidence strength ratings and gap indicators. Includes eval for unmapped assumptions and circular evidence chains.
Uncertainty Annotation Prompt for Plans
For plan generation outputs that need explicit uncertainty markup. Annotates each plan step with uncertainty type, source, and propagation risk. Produces an uncertainty-augmented plan suitable for risk-aware execution engines. Includes checks for unannotated irreversible steps.
Assumption Conflict Detection Prompt Template
For multi-source planning where assumptions may contradict. Detects logical conflicts between assumptions from different sources or agents. Produces a conflict report with resolution suggestions and impacted plan steps. Includes eval for missed contradictions and false conflict flags.
Evidence Gap Analysis Prompt Template
For planning modules identifying what evidence is missing before execution can proceed safely. Analyzes the plan against available evidence and identifies critical gaps. Produces a prioritized evidence collection plan with specific information requests. Includes harness for gap criticality scoring.
Assumption Validation Prompt Against Evidence
For runtime verification that assumptions still hold as new evidence arrives. Tests each active assumption against fresh evidence and produces a validation status. Includes decay detection for assumptions that were previously validated but may have aged out. Includes eval for false validation and missed invalidation.
Risk-of-Silence Assessment Prompt Template
For agents operating with incomplete user input where missing information could cause confident errors. Assesses what critical information the user has not provided and the risk of proceeding without it. Produces a risk score and recommended action. Includes checks for over-asking and under-detection.
Assumption Drift Monitoring Prompt Template
For long-running agent workflows where assumptions may degrade over time. Compares current state against baseline assumptions and detects drift. Produces a drift report with affected plan steps and recommended replanning triggers. Includes harness for drift threshold configuration and false-alarm tuning.
Confidence Score Assignment Prompt Template
For agents that need calibrated confidence estimates on individual claims or steps. Assigns confidence scores with explicit justification and known limitations. Produces confidence-annotated outputs suitable for downstream threshold decisions. Includes calibration eval against actual outcomes.
Assumption-Based Stop Condition Prompt Template
For autonomous agents that need to know when to stop and escalate. Defines stop conditions based on assumption violations, confidence drops, or evidence contradictions. Produces executable stop rules with clear trigger conditions. Includes eval for premature stopping and failure to stop when warranted.
Uncertainty Escalation Trigger Prompt Template
For supervised autonomy workflows where uncertainty must trigger human review. Defines escalation criteria based on confidence thresholds, action irreversibility, and assumption criticality. Produces an escalation decision with context package for human reviewer. Includes harness for escalation latency and false-escalation tracking.
Assumption Dependency Graph Generation Prompt Template
For complex plans where assumptions form dependency chains. Extracts assumption dependencies and produces a directed graph showing propagation paths. Identifies which plan steps are impacted if a root assumption fails. Includes eval for missing dependencies and circular references.
Evidence Sufficiency Evaluation Prompt Template
For agents deciding whether collected evidence is adequate to close an assumption. Evaluates evidence completeness, source reliability, and contradiction status. Produces a sufficiency determination with remaining gaps. Includes harness for sufficiency criteria configuration per assumption type.
Assumption Communication Prompt for Handoffs
For multi-agent systems and human handoffs where assumptions must transfer with context. Generates a structured assumption summary suitable for the receiving agent or human. Produces a handoff package with critical assumptions, confidence levels, and open questions. Includes eval for assumption loss during transfer.
Uncertainty-Aware Human Handoff Prompt Template
For agents that must hand off to humans when uncertainty exceeds operational thresholds. Generates a structured handoff with current state, specific uncertainties, what has been tried, and what the human needs to decide. Produces a concise handoff context that minimizes human ramp-up time. Includes eval for handoff completeness and unnecessary handoffs.
Multi-Step Reasoning and Chain-of-Thought Planning
Prompt playbooks for structuring the model's internal reasoning across multiple planning steps, including chain-of-thought, tree-of-thought, and reflection patterns. Useful for complex planning scenarios where single-pass plan generation produces shallow or inconsistent results.
Chain-of-Thought Planning Prompt Template for Complex Workflows
For agent orchestrators that need to decompose a high-level objective into a reasoned, step-by-step plan. Produces a plan with explicit reasoning traces for each step. Includes eval checks for logical consistency, step necessity, and goal alignment.
Tree-of-Thought Exploration Prompt for Architecture Decisions
For technical decision-makers evaluating multiple solution paths. Generates a branching exploration of options with pros, cons, and a recommended path. Includes harness logic for pruning, depth control, and comparing leaf-node evaluations.
Reflection and Self-Critique Prompt for Agent Plans
For planning modules that must validate a generated plan before execution. Produces a structured critique identifying gaps, infeasible steps, and missing dependencies. Includes eval criteria for critique depth and actionable revision suggestions.
Chain-of-Thought Root Cause Analysis Prompt Template
For SRE and on-call engineers investigating production incidents. Produces a step-by-step causal chain from symptom to root cause, with evidence requirements at each link. Includes validation checks for logical leaps and unverified assumptions.
Tree-of-Thought Incident Remediation Path Prompt
For incident commanders evaluating multiple fix strategies under uncertainty. Generates a decision tree of remediation options with risk, blast radius, and rollback feasibility for each branch. Includes eval for option completeness and risk coverage.
Reflection Prompt for Code Review Agent Summaries
For AI coding agents that must self-critique their own pull request reviews. Produces a second-pass analysis that catches missed issues, overstatements, and false positives. Includes harness for comparing initial and reflected review quality.
Chain-of-Thought Query Optimization Prompt
For database engineers and ORM-heavy codebases. Produces a reasoned analysis of a slow query, walking through execution plan interpretation, index recommendations, and rewrite options. Includes eval for correctness against schema and cardinality assumptions.
Tree-of-Thought Data Model Design Prompt
For data architects choosing between normalization strategies, access patterns, and consistency models. Generates a branching comparison of schema designs with trade-off analysis. Includes harness for constraint satisfaction and anti-pattern detection.
Reflection Prompt for Test Coverage Gap Analysis
For QA engineers reviewing a test suite against a code change. Produces a self-critique of an initial coverage assessment, identifying untested edge cases and over-tested paths. Includes eval for missed boundary conditions and equivalence class gaps.
Chain-of-Thought API Integration Flow Design Prompt
For integration engineers designing multi-service orchestration. Produces a step-by-step reasoning trace for request sequencing, error propagation, and state management across API calls. Includes validation for idempotency, timeouts, and partial failure handling.
Tree-of-Thought Authentication Pattern Selection Prompt
For security engineers choosing between OAuth flows, token strategies, and session models. Generates a decision tree comparing patterns against threat model, user experience, and operational constraints. Includes eval for security property coverage.
Reflection Prompt for Postmortem Drafting
For incident analysts producing blameless postmortems. Takes a draft postmortem and produces a self-critique that identifies missing contributing factors, weak action items, and hindsight bias. Includes harness for timeline consistency and causal completeness.
Chain-of-Thought Log Analysis and Anomaly Detection Prompt
For operators investigating anomalous system behavior from log streams. Produces a reasoned walkthrough connecting log patterns to hypotheses, ruling out false leads, and converging on a root cause. Includes eval for evidence-to-conclusion traceability.
Tree-of-Thought Performance Optimization Prompt
For performance engineers exploring optimization strategies across caching, query changes, and infrastructure scaling. Generates a branching analysis of approaches with cost, risk, and expected impact estimates. Includes harness for bottleneck validation.
Reflection Prompt for Code Smell Identification
For static analysis agents that must double-check their own findings. Produces a second-pass review of detected code smells, filtering false positives and prioritizing by impact. Includes eval for smell classification accuracy and remediation actionability.
Chain-of-Thought Monolith Decomposition Prompt
For architects planning service extraction from a monolith. Produces a reasoned sequence of extraction steps with coupling analysis, data migration checkpoints, and rollback points. Includes validation for bounded context integrity and interface stability.
Tree-of-Thought Capacity Planning Prompt
For infrastructure planners evaluating scaling strategies under demand uncertainty. Generates a branching model of capacity options with cost, latency, and failure mode analysis. Includes harness for assumption sensitivity and worst-case scenario coverage.
Reflection Prompt for Security Architecture Review
For security reviewers validating a system design. Produces a self-critique of an initial threat assessment, catching overlooked attack surfaces and weak controls. Includes eval for threat model coverage and control effectiveness reasoning.
Chain-of-Thought Dependency Resolution Prompt for Package Updates
For developers managing complex dependency upgrades. Produces a reasoned walkthrough of transitive dependency conflicts, breaking changes, and safe upgrade ordering. Includes validation for version constraint satisfaction and regression risk.
Tree-of-Thought Zero-Downtime Release Strategy Prompt
For release engineers planning database migrations with deployment sequencing. Generates a decision tree of release patterns comparing blue-green, canary, and feature-flag approaches. Includes eval for data compatibility windows and rollback safety.
Plan Monitoring and Notification
Prompt playbooks for defining what conditions should trigger a status update, alert, or human notification during plan execution. Useful for supervised autonomy and long-running agent workflows where operators need visibility into progress without polling or reading full traces.
Agent Execution Heartbeat Prompt Template
For operators of long-running agent workflows. Produces a structured status summary at configurable intervals, including current step, progress percentage, last completed milestone, and any blocked dependencies. Includes eval checks for stale timestamps and missing step identifiers.
Human Escalation Trigger Prompt Template
For supervised autonomy systems. Decides whether current execution state requires human intervention based on configurable policies: confidence thresholds, risk levels, authorization boundaries, or unexpected outputs. Produces an escalation decision with structured reasoning and a proposed handoff summary.
Stuck Agent Detection and Alert Prompt
For agent runtime monitors. Analyzes execution traces to detect stall patterns: repeated tool calls without progress, looped reasoning, or steps exceeding time budgets. Produces a stuck-state classification with evidence and a recommended recovery action.
Deadline Miss Prediction Notification Prompt
For SLA-bound agent workflows. Compares current progress rate against remaining steps and deadline constraints to predict completion feasibility. Produces a risk assessment with projected completion time, confidence interval, and recommended mitigation.
Cost Overrun Alert Prompt for Token Usage
For teams managing LLM cost budgets. Monitors cumulative token consumption against plan estimates and triggers alerts when projected total exceeds thresholds. Produces a cost variance report with per-step breakdown and optimization recommendations.
Tool Call Failure Alert Prompt
For agent operators tracking external dependency health. Classifies tool call failures by severity and recoverability: transient errors, authentication failures, schema mismatches, or service unavailability. Produces a failure classification with recommended retry or escalation path.
Plan Deviation Detection Prompt
For execution monitors comparing actual agent behavior against the original plan. Detects when the agent skips steps, reorders execution, adds unplanned actions, or changes output targets. Produces a deviation report with severity rating and plan-sync recommendation.
Context Window Overflow Warning Prompt
For long-running agent sessions approaching token limits. Estimates remaining context budget and identifies which content to summarize, evict, or preserve. Produces a context budget report with prioritization recommendations to prevent mid-execution truncation.
Dependency Failure Cascade Notification Prompt
For multi-step workflows with prerequisite chains. When an upstream step fails, analyzes which downstream steps are now blocked, invalidated, or need re-execution. Produces an impact map with affected step IDs and recommended replanning scope.
Human Approval Request Prompt Template
For agents reaching gated actions requiring authorization. Produces a structured approval request including the proposed action, its consequences, alternatives considered, and risk assessment. Designed for integration with review queues and approval workflows.
Task Completion Summary Prompt for Operators
For agents finishing a multi-step workflow. Produces a concise operator summary covering what was accomplished, any deviations from plan, unresolved items, artifacts produced, and recommended follow-up actions. Includes completeness checks against original objectives.
Partial Failure Status Update Prompt
For workflows where some steps succeeded and others failed. Produces a structured status distinguishing completed work from failed steps, preserving partial results, and recommending whether to resume, rollback, or escalate. Includes artifact inventory for completed steps.
Retry Exhausted Escalation Prompt
For agents that have exhausted configured retry attempts on a failing step. Produces an escalation summary including the original objective, failure history across retries, error patterns observed, and a recommendation to abort, skip, or request human intervention.
Silent Failure Heuristic Check Prompt
For detecting failures that produce valid-looking but incorrect outputs. Analyzes step outputs against expected patterns, value ranges, and consistency rules to flag silent failures like empty results, schema-compliant garbage, or confidence drops without explicit errors.
Output Drift Detection Alert Prompt
For monitoring agents over extended runs. Compares recent outputs against baseline expectations to detect gradual quality degradation, format drift, or behavioral shifts. Produces a drift report with trend indicators and recommended intervention thresholds.
Checkpoint Reached Notification Prompt
For agents configured with explicit save points. Produces a checkpoint summary when a milestone is reached, including completed work, current state snapshot, remaining plan, and any assumption changes since last checkpoint. Enables resume-after-interruption workflows.
Loop Detection and Repetition Alert Prompt
For runtime monitors analyzing agent action patterns. Detects when the agent repeats the same tool calls, reasoning patterns, or output structures without making progress. Produces a loop classification with repetition count, pattern description, and break recommendation.
Goal Abandonment Decision Prompt
For agents that cannot complete an objective after exhausting all recovery paths. Evaluates whether to formally abandon the goal, produce a partial deliverable, or escalate for re-scoping. Produces an abandonment recommendation with rationale and artifact preservation instructions.
Post-Mortem Trigger Notification Prompt
For workflows that complete with failures, escalations, or significant deviations. Produces a structured post-mortem trigger including incident summary, timeline of key events, root cause hypotheses, and recommended data to preserve for the full post-mortem process.
Plan Archiving and Reuse
Prompt playbooks for storing completed plans, extracting reusable patterns, and adapting prior plans to similar goals. Useful for teams building plan libraries and template systems because plan reuse reduces latency, cost, and variance across repeated workflow types.
Completed Plan Serialization Prompt Template
For agent orchestrators that need to persist finished plans as structured JSON for archival. Produces a normalized plan record with steps, dependencies, tool calls, and outcomes. Includes eval checks for schema compliance and completeness of the serialized state.
Plan Metadata Tagging Prompt for Searchable Libraries
For teams building searchable plan libraries. Generates structured metadata tags—goal type, domain, tools used, constraints, success criteria—from a completed plan. Includes eval checks for tag consistency and retrieval recall against known queries.
Plan Similarity Matching Prompt for Reuse Candidates
For plan library systems that need to find the closest archived plan to a new goal. Produces a ranked list of candidate plans with similarity scores and rationale. Includes eval checks for ranking quality and false-positive rejection.
Plan Adaptation Prompt for Similar Goal Transfer
For agent runtimes that need to adapt an archived plan to a new but similar objective. Produces a modified plan with updated steps, constraints, and tool mappings while preserving the proven structure. Includes eval checks for constraint preservation and step relevance.
Reusable Plan Pattern Extraction Prompt
For teams building plan template libraries from historical runs. Extracts the abstract workflow pattern from a concrete plan, removing instance-specific details while preserving the decision logic and step structure. Includes eval checks for pattern generality and over-abstraction.
Plan Template Generalization Prompt from Completed Runs
For converting one or more completed plan instances into a parameterized template. Produces a template with declared variables, default values, and insertion points. Includes eval checks for variable completeness and template instantiation validity.
Plan Confidence Scoring Prompt for Reuse Suitability
For plan reuse systems that need to assess whether an archived plan is safe to adapt. Produces a confidence score with factor breakdown—goal similarity, tool availability, constraint match, known failure modes. Includes eval checks for score calibration against actual execution outcomes.
Plan Gap Analysis Prompt Before Reuse
For agent pre-execution checks that compare a candidate reused plan against the current goal and environment. Produces a structured gap report identifying missing steps, incompatible tools, and unsatisfied preconditions. Includes eval checks for gap recall and false-alarm rate.
Plan Constraint Substitution Prompt for New Contexts
For adapting a plan template when the target environment has different constraints—rate limits, permissions, data schemas, or latency budgets. Produces a revised plan with constraint-aware step modifications. Includes eval checks for constraint violation detection.
Plan Tool Mapping Prompt for Different Agent Environments
For migrating a plan between agent frameworks or tool ecosystems. Produces a tool substitution map and any required step rewrites when the original tools are unavailable. Includes eval checks for capability equivalence and missing-tool detection.
Plan Parameterization Prompt for Reusable Templates
For converting a concrete plan into a template with explicit parameters, types, and constraints. Produces a parameterized plan definition ready for a template engine. Includes eval checks for parameter coverage and type safety during instantiation.
Plan Failure Mode Extraction Prompt from Past Runs
For improving plan reuse safety by extracting known failure modes from execution logs. Produces a structured catalog of failure modes, triggers, and recommended mitigations associated with a plan. Includes eval checks for failure mode recall against incident reports.
Plan Dependency Update Prompt for Reused Workflows
For updating a reused plan when upstream APIs, tool versions, or data schemas have changed. Produces a revised plan with updated dependency references and compatibility notes. Includes eval checks for stale reference detection.
Plan Composition Prompt from Multiple Archived Plans
For building a new plan by combining steps from multiple archived plans. Produces a merged plan with resolved ordering, deduplicated steps, and conflict notes. Includes eval checks for step redundancy and ordering consistency.
Plan Conflict Resolution Prompt for Merged Templates
For resolving conflicts when two plan templates specify contradictory steps, constraints, or tool choices. Produces a reconciled plan with conflict annotations and resolution rationale. Includes eval checks for unresolved conflict detection.
Plan Versioning Prompt for Template Evolution
For managing plan template changes over time with clear version deltas and migration notes. Produces a versioned template record with change descriptions and backward-compatibility flags. Includes eval checks for version lineage integrity.
Plan Deprecation Prompt for Outdated Workflows
For marking archived plans as deprecated when they reference retired tools, APIs, or unsafe patterns. Produces a deprecation record with replacement suggestions and migration guidance. Includes eval checks for false deprecation and missing replacement coverage.
Plan Checkpoint Reuse Prompt for Long-Running Workflows
For resuming a long-running workflow from an archived checkpoint rather than restarting. Produces a state-reconstruction plan that validates checkpoint integrity and identifies any stale context. Includes eval checks for state consistency after resume.
Plan Error Recovery Pattern Extraction Prompt
For extracting reusable error recovery strategies from plans that successfully handled failures. Produces a recovery pattern catalog with trigger conditions, recovery steps, and rollback logic. Includes eval checks for pattern completeness against incident timelines.
Plan Audit Trail Prompt for Reuse Governance
For generating an audit record when a plan is reused, adapted, or composed. Produces a traceable record linking the new plan to source plans, adaptations made, and approval decisions. Includes eval checks for audit completeness and lineage traceability.
Partnered with leading AI, data, and software stack.
How We Work
Custom AI workflows for your Business
One-fit-all AI don't work for modern businesses. At Inferensys, we aim to understand your business & custom requirements; which we use to define most efficient agentic workflows, the data, and the tools for your business.
01
Review the use case
We understand the task, the users, and where AI can actually help.
Read more02
Pick the right approach
We define what needs search, automation, or product integration.
Read more03
Build the first useful version
We implement the part that proves the value first.
Read more04
Improve from there
We add the checks and visibility needed to keep it useful.
Read moreThe first call is a practical review of your use case and the right next step.
Talk to Us