postmortem-writing — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited postmortem-writing (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Use this skill when authoring or reviewing an incident report (IR) for the Lorekeeper project. Every incident must produce durable process improvements — not just a fix but encoding the lesson into review patterns and checklists.
mainEvery IR MUST deliver all of:
| # | Deliverable | Location | Why |
|---|---|---|---|
| 1 | docs/incidents/IR-NNN-description.md | Repo docs/incidents/ | Permanent record |
| 2 | BLOCKER pattern(s) added to lorekeeper-code-reviewer | .hermes/skills/lorekeeper-code-reviewer/SKILL.md | Prevents same bug class passing review |
| 3 | Checklist items added to github-code-review-checklist | Skills directory | Catches in general review too |
| 4 | Engineering docs (if the pattern needs explanation) | docs/engineering/ | Reference for future readers |
| 5 | Action items tracked to completion | IR action items table | Closure |
---
**Status:** [Resolved | In Progress | Monitoring]
**Severity:** [P0/P1/P2]
**Date:** YYYY-MM-DD
**Author:** [Name]
**Fix PR:** #[number]
---#### 1. Summary
One paragraph: what happened, why it was bad, what fixed it. Include the key insight in the first sentence.
#### 2. Timeline
Table of events in chronological order: time + event. Include first symptom, investigation start, fix identified, fix merged, recovery confirmed.
#### 3. Root Causes — Causal Chain
This is the most important section. Model the incident as a chain of independent links, where each link is survivable alone but together they create the failure.
Format each link as:
### 🥇 Link N: [Title]
[Technical explanation — what the link is]
**Why this caused today's crash:** [How this specific link contributed]
**Evidence:** [Log lines, metrics, code diffs, process output]After the chain, show how the fixes break specific links:
| Fix | Which link it breaks | Why it suffices alone |
| ----------------------- | -------------------- | --------------------- |
| `busy_timeout=5000` | Link 2 | ... |#### 4. Impact
#### 5. Detection
How the incident was first noticed (user report, alert, dashboard). Include evidence commands (e.g. ps aux, curl checks).
#### 6. Fixes Applied
Group by PR. For each PR, list changes as a table: what changed, file(s), what it does.
Separate emergency fixes (stop the bleeding) from structural fixes (eliminate the class).
#### 7. Action Items
Table: # | Action | Owner | Status
Every action item must be concrete and assigned. No "investigate further" — each item has a clear done condition.
Must include:
#### 8. Lessons Learned
Numbered list of principles extracted from this incident. These go into the causal chain analysis section and become future review heuristics.
#### 9. Related
Links to: PRs, commits, related skills, engineering docs, previous IRs.
The causal chain is the key insight from IR-002. Rules:
commit() that passed all tests).Optionally include an ASCII causal chain diagram:
[Link 1 crash] ─→ [Link 2 instant error]
│ │
│ └──→ [User-visible symptom]
│
└──→ [Link 3 infinite retries] ─→ [Link 4 zombies amplify]
│
└──→ [Every init hits contention]After the IR is written, verify:
docs/incidents/ with sequential numberinglorekeeper-code-reviewer (with concrete code examples)github-code-review-checklistdocs/engineering/~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.