# Debug Reproduced

> User confirms the bug was reproduced. Analyze logs and iterate on hypotheses.

- Skill: `mikecfisher/debug-reproduced` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add mikecfisher/debug-reproduced`
- Raw SKILL.md: https://api.skillmd.com/api/skills/mikecfisher/debug-reproduced/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: mikecfisher (https://skillmd.com/u/mikecfisher)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/mikecfisher/debug-reproduced

---


# Debug Reproduced

User has reproduced the bug. Analyze log evidence and iterate.

## User Notes

$ARGUMENTS

## Workflow

### Step 1: Analyze Logs

Run the analyzer:

```bash
bun ./scripts/analyze-logs.mjs
```

Or read raw logs:

```bash
cat .debug/debug.log
```

### Step 2: Evaluate Hypotheses

For each hypothesis, assign one verdict:

- **CONFIRMED**: Log evidence proves this is the cause
- **REJECTED**: Log evidence rules this out
- **INCONCLUSIVE**: Need more instrumentation

**Always cite specific log entries as evidence:**

> Hypothesis A: **CONFIRMED**
>
> - Log at 14:32:05.123 shows `items` is `undefined` at function entry
> - Log at 14:32:05.156 shows error "Cannot read property 'length' of undefined"
> - Root cause: `order.items` not provided by caller
>
> Hypothesis B: **REJECTED**
>
> - Logs show `loadUser` completed at 14:32:05.100
> - `renderProfile` started at 14:32:05.200
> - Order is correct, no race condition

### Step 3: Take Action

**If hypothesis CONFIRMED:**

1. Implement the fix
2. **Keep instrumentation in place** (do NOT remove logs yet)
3. Provide new reproduction steps to verify fix
4. After user reproduces, **compare before/after logs** — cite specific entries showing the fix worked

**If INCONCLUSIVE:**

1. Add more instrumentation to narrow down
2. Provide new reproduction steps

**If ALL hypotheses REJECTED:**

1. Generate NEW hypotheses from **different subsystems**
2. Add instrumentation for new hypotheses
3. Provide new reproduction steps

### Step 4: Next Iteration

Provide exact reproduction steps:

<reproduction_steps>

1. [Step 1]
2. [Step 2]
   ...
   </reproduction_steps>

End with:

> Follow these steps, then run `/debug-reproduced` when done, or `/debug-fixed` if the issue is resolved.

## Critical Rules

1. **NEVER fix without 100% confidence with log proof**
2. **NEVER remove instrumentation before verified fix**
3. **ALWAYS cite specific log entries** — "Log at 14:32:05 shows X was null"
4. **Iteration is expected** — Failed fixes mean new hypotheses, not giving up
5. **3 failed fixes = step back** — Reconsider assumptions, look at different subsystems
6. **Taking longer is better** — Thorough debugging beats quick guesses

## Note on Log Accumulation

Logs accumulate across iterations (not auto-cleared). This allows comparing behavior across reproduction cycles. Run `bun ./scripts/clear-logs.mjs` if you need a fresh start.