# Incident Learning Loop

> Convert incidents into durable standards, safeguards, templates, and design rules that reduce recurrence across the system.

- **Type:** Skill
- **Install:** `agentstack add skill-caiaffa-claude-code-ultimate-engineering-system-incident-learning-loop`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [caiaffa](https://agentstack.voostack.com/s/caiaffa)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [caiaffa](https://github.com/caiaffa)
- **Source:** https://github.com/caiaffa/claude-code-ultimate-engineering-system/tree/main/skills/incident-learning-loop

## Install

```sh
agentstack add skill-caiaffa-claude-code-ultimate-engineering-system-incident-learning-loop
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Mission
Ensure every serious incident improves the engineering system, not just the affected service.

# When to use
- Closing a postmortem.
- Deciding what should become a standard.
- Updating templates, checklists, or runbooks after incidents.
- Propagating lessons beyond the local failure.

# Handoff
- **Receives from:** postmortem-reviewer (after postmortem analysis) or deep-root-cause-investigator.
- **Hands off to:** principal-engineer (for standard updates), operational-excellence-enforcer (for runbook/alert updates).

# The learning extraction process
For every incident, ask:

1. **What class of failure was this?** (not "what happened" but "what category")
   - Examples: missing idempotency, schema migration lock, stale cache read, missing circuit breaker
2. **What invariant was violated?** (check against SYSTEM_INVARIANTS.md)
3. **What standard should change?** (PROJECT_CONVENTIONS.md, DECISION_RULES.md, etc.)
4. **What template or checklist should be updated?** (SERVICE_SCORECARD.md, DEFINITION_OF_DONE.md, etc.)
5. **What other services have this same exposure?** (lateral scan)
6. **What review requirement should be added?** (new item in code review, ADR review, etc.)

# Red flags — learning is too shallow if
- Lesson stays local to one team or service.
- Fix applied to code but no standard or process changes.
- Runbook improved but the design that caused the incident is unchanged.
- Same class of failure is still possible in 3 other services.
- Action items are all "add monitoring" without prevention.

# Output format
1. **Class of failure** (categorized)
2. **Systemic lesson** (what the org should learn)
3. **Standards to update** (specific files and changes)
4. **Templates/checklists to add or modify** (specific)
5. **Other exposed services** (lateral risk scan)
6. **Wider rollout plan** (how to propagate the fix beyond this service)

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [caiaffa](https://github.com/caiaffa)
- **Source:** [caiaffa/claude-code-ultimate-engineering-system](https://github.com/caiaffa/claude-code-ultimate-engineering-system)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-caiaffa-claude-code-ultimate-engineering-system-incident-learning-loop
- Seller: https://agentstack.voostack.com/s/caiaffa
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.