Conducting Post-Mortem

Conducting Post-Mortem

Analyzes the current conversation to identify lessons learned and proposes specific updates to CLAUDE.md to prevent future issues.

When Invoked

This skill automatically reviews the current conversation context to extract actionable insights.

Arguments

None required - Automatically analyzes the current conversation history and context.

Optional: Path to a completed task file

• If provided: Read that file as additional context alongside conversation history to improve gotcha identification quality
• If omitted: Rely on conversation history alone

Process

Now that we are done with the task:

1. Review our conversation history. If a task file path was provided, read that file as additional context.
2. Identify up to 3 "Unknown Unknowns" or "Gotchas" we encountered (e.g., a library quirk, a missing permission, a tricky type error). Rank them by potential impact on future sessions — most impactful first.
3. For each gotcha not already in ./CLAUDE.md, propose specific markdown content: describe the problem, the solution, and suggest where in CLAUDE.md to add it.

Output Format

Format the output as a numbered list, one entry per proposed gotcha, ranked by impact (most impactful first):

## PROPOSED CLAUDE.MD UPDATES

### 1. [Gotcha Title] *(highest impact)*
**Problem**: [What went wrong or was surprising]
**Solution**: [The fix or pattern to remember]
**Suggested location**: CLAUDE.md → [section path]

### 2. [Gotcha Title]
...

### 3. [Gotcha Title] *(lowest impact)*
...

Include only gotchas not already covered in CLAUDE.md. If fewer than 3 novel gotchas were found, include only those found.

Examples

Usage:

/conducting-post-mortem

Sample Output:

## PROPOSED CLAUDE.MD UPDATE

### Gotcha: Supabase Edge Functions Require Explicit CORS Headers

**Problem**: Edge functions were returning 200 but browser blocked responses due to missing CORS headers.

**Solution**: Always include in Edge Function responses:
```typescript
return new Response(JSON.stringify(data), {
  headers: {
    'Content-Type': 'application/json',
    'Access-Control-Allow-Origin': '*',
    'Access-Control-Allow-Headers': 'authorization, x-client-info, apikey, content-type',
  },
})

Where to Add: CLAUDE.md → Supabase → Edge Functions section

## Error Handling

- **No CLAUDE.md**: WARN — still propose the update content, note that CLAUDE.md needs to be created first
- **No meaningful gotchas found**: Report "No novel gotchas identified in this session" and stop