agent-introspection-debugging
Structured self-debugging workflow for AI agent failures using capture, diagnosis, contained recovery, and introspection reports. Use when an agent run fails and you need a reproducible diagnosis instead of a retry.
- License 许可
- MIT
- 所属仓库
- shaikhsarib/Deep
- 最近更新
- 2026-11-07
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- agent-introspection-debugging
- license
- MIT
- description
- Structured self-debugging workflow for AI agent failures using capture, diagnosis, contained recovery, and introspection reports. Use when an agent run fails and you need a reproducible diagnosis instead of a retry.
02 资源文件 1
- 参考文档agents/openai.yaml302 B
03 SKILL.md 正文
Agent Introspection Debugging
Use this skill when an agent run is failing repeatedly, consuming tokens without progress, looping on the same tools, or drifting away from the intended task.
This is a workflow skill, not a hidden runtime. It teaches the agent to debug itself systematically before escalating to a human.
When to Activate
- Maximum tool call / loop-limit failures
- Repeated retries with no forward progress
- Context growth or prompt drift that starts degrading output quality
- File-system or environment state mismatch between expectation and reality
- Tool failures that are likely recoverable with diagnosis and a smaller corrective action
Scope Boundaries
Activate this skill for:
- capturing failure state before retrying blindly
- diagnosing common agent-specific failure patterns
- applying contained recovery actions
- producing a structured human-readable debug report
Do not use this skill as the primary source for:
- feature verification after code changes; use
verification-loop - framework-specific debugging when a narrower ECC skill already exists
- runtime promises the current harness cannot enforce automatically
Four-Phase Loop
Phase 1: Failure Capture
Before trying to recover, record the failure precisely.
Capture:
- error type, message, and stack trace when available
- last meaningful tool call sequence
- what the agent was trying to do
- current context pressure: repeated prompts, oversized pasted logs, duplicated plans, or runaway notes
- current environment assumptions: cwd, branch, relevant service state, expected files
Minimum capture template:
## Failure Capture
- Session / task:
- Goal in progress:
- Error:
- Last successful step:
- Last failed tool / command:
- Repeated pattern seen:
- Environment assumptions to verify:
Phase 2: Root-Cause Diagnosis
Match the failure to a known pattern before changing anything.
| Pattern | Likely Cause | Check |
|---|---|---|
| Maximum tool calls / repeated same command | loop or no-exit observer path | inspect the last N tool calls for repetition |
| Context overflow / degraded reasoning | unbounded notes, repeated plans, oversized logs | inspect recent context for duplication and low-signal bulk |
ECONNREFUSED / timeout |
service unavailable or wrong port | verify service health, URL, and port assumptions |
429 / quota exhaustion |
retry storm or missing backoff | count repeated calls and inspect retry spacing |
| file missing after write / stale diff | race, wrong cwd, or branch drift | re-check path, cwd, git status, and actual file existence |
| tests still failing after “fix” | wrong hypothesis | isolate the exact failing test and re-derive the bug |
Diagnosis questions:
- is this a logic failure, state failure, environment failure, or policy failure?
- did the agent lose the real objective and start optimizing the wrong subtask?
- is the failure deterministic or transient?
- what is the smallest reversible action that would validate the diagnosis?
Phase 3: Contained Recovery
Recover with the smallest action that changes the diagnosis surface.
Safe recovery actions:
- stop repeated retries and restate the hypothesis
- trim low-signal context and keep only the active goal, blockers, and evidence
- re-check the actual filesystem / branch / process state
- narrow the task to one failing command, one file, or one test
- switch from speculative reasoning to direct observation
- escalate to a human when the failure is high-risk or externally blocked
Do not claim unsupported auto-healing actions like “reset agent state” or “update harness config” unless you are actually doing them through real tools in the current environment.
Contained recovery checklist:
## Recovery Action
- Diagnosis chosen:
- Smallest action taken:
- Why this is safe:
- What evidence would prove the fix worked:
Phase 4: Introspection Report
End with a report that makes the recovery legible to the next agent or human.
## Agent Self-Debug Report
- Session / task:
- Failure:
- Root cause:
- Recovery action:
- Result: success | partial | blocked
- Token / time burn risk:
- Follow-up needed:
- Preventive change to encode later:
Recovery Heuristics
Prefer these interventions in order:
- Restate the real objective in one sentence.
- Verify the world state instead of trusting memory.
- Shrink the failing scope.
- Run one discriminating check.
- Only then retry.
Bad pattern:
- retrying the same action three times with slightly different wording
Good pattern:
- capture failure
- classify the pattern
- run one direct check
- change the plan only if the check supports it
Integration with ECC
- Use
verification-loopafter recovery if code was changed. - Use
continuous-learning-v2when the failure pattern is worth turning into an instinct or later skill. - Use
councilwhen the issue is not technical failure but decision ambiguity. - Use
workspace-surface-auditif the failure came from conflicting local state or repo drift.
Output Standard
When this skill is active, do not end with “I fixed it” alone.
Always provide:
- the failure pattern
- the root-cause hypothesis
- the recovery action
- the evidence that the situation is now better or still blocked
04 同仓库更多 Skills
energy-procurement
Procure electricity and natural gas for commercial and industrial facilities: tariff and rate-schedule optimization, demand-charge mitigation, supplier RFPs, fixed/index/block-and-index hedging, renewable PPA and REC evaluation, and sustainability reporting. Use when procuring energy, optimizing utility tariffs, managing demand charges, evaluating PPAs, or building energy budgets and hedge strategies.
data-throughput-accelerator
Diagnose and accelerate large data movement — ingestion, backfill, export, ETL, warehouse loading, manifest catch-up, and table synchronization — by isolating the true bottleneck, benchmarking variants, and codifying the fastest path with a hard accounting block proving rows and timestamps cohere. Use when a pipeline or backfill is too slow and must get faster without losing data correctness.
customs-trade-compliance
Codified customs and trade compliance expertise — HS/HTS tariff classification with GRI rules, commercial invoices and entry documentation, Incoterms 2020, FTA qualification and duty optimization (USMCA, RCEP, FTZs, drawback), denied-party screening, and penalty mitigation across US, EU, UK, and APAC jurisdictions. Use when classifying goods, preparing import/export documentation, screening restricted parties, responding to customs audits or CF-28/penalty notices, or optimizing duties.
carrier-relationship-management
Manage truckload, LTL, and intermodal carrier portfolios: sourcing and FMCSA vetting, freight rate and fuel-surcharge negotiation, RFPs and routing guides, carrier scorecards, allocation, and renewals. Use when onboarding carriers, running freight RFPs, negotiating rates, evaluating carrier performance, reallocating freight, or building freight strategy.
benchmark
Measure performance baselines and detect regressions across browser Core Web Vitals (LCP, INP, CLS, page weight), API endpoint latency percentiles, and build/test feedback times, with before/after comparison stored in git-tracked .ecc/benchmarks JSON. Use when checking page speed, responding to 'it feels slow' reports, verifying launch performance targets, or comparing stack alternatives.
benchmark-optimization-loop
Convert 'make it faster' requests into a bounded measured optimization loop — baseline first, generate one-hypothesis variants, benchmark each against a correctness gate, and promote the fastest safe variant with reproducible commands. Use when asked to speed something up, try many variants, run recursive optimization, benchmark latency/throughput/cost, or pick the best implementation by repeated measured tests.