# Agent Introspection Debugging

> Structured self-debugging workflow for AI agent failures using capture, diagnosis, contained recovery, and introspection reports.

- Skill: `yevhens-hue/agent-introspection-debugging` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add yevhens-hue/agent-introspection-debugging`
- Raw SKILL.md: https://api.skillmd.com/api/skills/yevhens-hue/agent-introspection-debugging/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: yevhens-hue (https://skillmd.com/u/yevhens-hue)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/yevhens-hue/agent-introspection-debugging

---


# 🕵️ Agent Introspection & Failure Injection Debugging Skill

This skill governs structured self-debugging and failure injection testing for AI agents and distributed systems.

---

## 🛠️ Failure Injection Protocol

1. **Synthetic Failure Scenarios:**
   - **Scenario A (Database Timeout):** Simulate PostgreSQL / Redis connection drop. Verify fallback memory cache triggers cleanly.
   - **Scenario B (Rate Limit 429):** Inject simulated LLM HTTP 429 response. Verify exponential backoff retry logic.
   - **Scenario C (Missing Credentials):** Run without API keys. Verify graceful failure log and administrator alert dispatch.

2. **System Recovery & Verification Map:**
   - Document **Verified Components** (tested empirically under failure assertions).
   - Document **Unverified Gaps** (edge cases requiring additional test coverage).