harness-diagnose
Skill48 starsChanged 6 months agoApache-2.0View on GitHub
Disciplined diagnosis workflow for failing gates, regressions, flaky behavior, performance regressions, and suspicious backend/runtime behavior. Use when the user reports a bug, broken invariant, failing command, failing test, or asks to diagnose/debug a problem.
In this repository13 other files it gives its agents.
- AGENTS.mdAGENTS.md
- backend-engineeringSkill
- contract-agentSkill
- harness-architecture-reviewSkill
- harness-grillSkill
- harness-tddSkill
- harness-zoom-outSkill
- plannerSkill
- platform-ops-agentSkill
- server-agentSkill
- service-agentSkill
- skill-authoringSkill
- worker-agentSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 6 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 11
Use this skill for bugs and regressions.
- Distinctive: line 22
Add regression evidence at the same semantic level.
- Very specific: line 23
Run path-scoped gates from agent/manifests/gate-matrix.yml.
- Distinctive: line 28
Prefer fast, deterministic, agent-runnable signals:
- Distinctive: line 38
If no credible loop can be built, stop and state what was tried and what artifact is needed.
- Distinctive: line 42
Every probe must test one hypothesis.
- Distinctive: line 43
Prefer debugger, targeted assertion, or narrow logs over broad logging.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Harness Diagnose
Use this skill for bugs and regressions.
Workflow skills guide process. Ownership skills still decide writable boundaries.
Phases
- Build a feedback loop.
- Reproduce the exact symptom.
- Generate 3-5 ranked falsifiable hypotheses.
- Instrument one boundary or variable at a time.
- Fix the smallest causal closed loop.
- Add regression evidence at the same semantic level.
- Run path-scoped gates from
agent/manifests/gate-matrix.yml. - Report evidence and residual uncertainty.
Feedback Loop Options
Prefer fast, deterministic, agent-runnable signals:
- failing unit, service, integration, replay, or contract test
- validator or gate command
- CLI invocation with fixture input and expected output
- HTTP request against a running server
- replayed event, outbox, or projector trace
- throwaway harness isolated to one service, server, worker, or repo-tool path
- stress loop for flakes or concurrency issues
If no credible loop can be built, stop and state what was tried and what artifact is needed.
Instrumentation Rules
- Every probe must test one hypothesis.
- Prefer debugger, targeted assertion, or narrow logs over broad logging.
- Tag temporary logs with a unique prefix and remove them before completion.
- For performance regressions, measure first and fix second.
Completion Report
State:
- violated invariant
- causal boundary: contract, service, server, worker, platform, tooling, docs, or cross-boundary
- repair made
- regression evidence added or why no correct seam exists
- gates run
- gates not run
- residual risk or uncertainty
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

