From debug-mode
User confirms the bug was reproduced. Analyze logs and iterate on hypotheses.
How this skill is triggered — by the user, by Claude, or both
Slash command
/debug-mode:debug-reproducedThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
User has reproduced the bug. Analyze log evidence and iterate.
User has reproduced the bug. Analyze log evidence and iterate.
$ARGUMENTS
Run the analyzer:
bun ./scripts/analyze-logs.mjs
Or read raw logs:
cat .debug/debug.log
For each hypothesis, assign one verdict:
Always cite specific log entries as evidence:
Hypothesis A: CONFIRMED
- Log at 14:32:05.123 shows
itemsisundefinedat function entry- Log at 14:32:05.156 shows error "Cannot read property 'length' of undefined"
- Root cause:
order.itemsnot provided by callerHypothesis B: REJECTED
- Logs show
loadUsercompleted at 14:32:05.100renderProfilestarted at 14:32:05.200- Order is correct, no race condition
If hypothesis CONFIRMED:
If INCONCLUSIVE:
If ALL hypotheses REJECTED:
Provide exact reproduction steps:
<reproduction_steps>
End with:
Follow these steps, then run
/debug-reproducedwhen done, or/debug-fixedif the issue is resolved.
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.
npx claudepluginhub mikecfisher/agent-debug-mode --plugin debug-modeScientific bug hunting using falsifiable hypotheses. Forms hypotheses, designs falsifying tests, eliminates candidates systematically, and logs the full investigation trail in a structured debug/ folder. TRIGGER when: user has a bug to investigate scientifically; user wants systematic root-cause analysis; user says "debug", "investigate", "root cause", "why is this failing"; user invokes /autoresearch:debug. DO NOT TRIGGER when: user wants to optimize a metric (use /autoresearch); user wants to fix a known error automatically (use /autoresearch:fix); user just wants a quick one-line answer about what a function does.
Hypothesis-driven debugging methodology with ranked hypotheses, git bisect strategies, instrumentation planning, and minimal reproduction design. For intermittent failures, unclear stacktraces, performance regressions, and non-obvious bugs requiring systematic investigation.
Forces a scientific-method loop (Observe → Hypothesize → Experiment → Conclude) to debug non-trivial bugs. Prevents guessing by ensuring evidence before fixes.