Harness Kit

skill

diagnose

Investigate, audit, triage, and fix. Systematic debugging, incident lifecycle, domain auditing, and issue logging. Feedback-loop-first protocol: reproduce or replay before root cause, pattern analysis, hypothesis test, and fix. Use for: any bug, test failure, production incident, error spikes, audit, triage, postmortem, "diagnose", "why is this broken", "debug this", "production down", "is production ok", "audit stripe", "log issues". Trigger: /diagnose.

/diagnose

Source: skills/diagnose/SKILL.md

What it does

Investigate, audit, triage, and fix. Systematic debugging, incident lifecycle, domain auditing, and issue logging. Feedback-loop-first protocol: reproduce or replay before root cause, pattern analysis, hypothesis test, and fix. Use for: any bug, test failure, production incident, error spikes, audit, triage, postmortem, "diagnose", "why is this broken", "debug this", "production down", "is production ok", "audit stripe", "log issues". Trigger: /diagnose.

Workflow role

Repair primitive in the Harness Kit operating loop.

Source contract preview

This generated excerpt gives readers the beginning of the live primitive contract before they jump to GitHub.

Find root cause. Fix it. Prove it works.

## Execution Stance

You are the executive orchestrator.
- Keep hypothesis ranking, root-cause proof, and fix selection on the lead model.
- Delegate bounded evidence gathering and implementation to focused subagents.
- Run parallel hypothesis probes when multiple plausible causes exist.
- For design-quality root causes, dispatch an ad-hoc **lens** critic
  (`harnesses/shared/references/lenses.md`) returning `finding · evidence
  file:line · impact` — name the lens, no static agent file.

## Delegation Floor

Delegation floor applies: probe the roster first; dispatch two or more
providers for substantive work; direct solo only for mechanical, emergency,
user-forbidden, or fewer-than-two-providers cases. See
`harnesses/shared/AGENTS.md` (Roster).

Local lane guidance: Use independent evidence or hypothesis lanes for competing root causes, reproduction, config/service checks, and proposed fix validation.

## Routing

| Intent | Sub-capability |
|--------|---------------|
| Debug a bug, test failure, unexpected behavior | This file (below) |
| Need a reproduction strategy | `references/feedback-loops.md` |
| Flaky test investigation | `references/flaky-test-investigation.md` |
| Incident lifecycle: triage, investigate, postmortem | `references/triage.md` |
| Domain audit: "audit stripe", "audit quality" | `references/audit.md` |
| Audit then fix highest priority issue | `references/fix.md` |
| Create GitHub issues from audit findings | `references/log-issues.md` |

If first argument matches a domain name (stripe, quality, etc.), route to `references/audit.md`.
If "triage", "incident", "postmortem", "production down" → `references/triage.md`.
If "flaky", "flake", "intermittent", "nondeterministic test" → `references/flaky-test-investigation.md`.
...

What to verify