harness
Skill1 starsChanged 37 days agoMITView on GitHub
Harness Institute entry skill for agentic coding. After Setup, three doors: bug, feature, campaign. Seven stations underneath. Use when the user says /harness, /harness bug, /harness feature, /harness campaign, run the harness skill, or asks to set up agentic habits in a repo. Also /harness loop and /harness continue. Explicit invocation only. Do not auto-start.
35 other projects use a version of this.
Adopted by
Most recently taken up by
- hanzhangzzz/agent-skills-zhtheir own version, first seen Oct 29, 2026
- anlmalabs/skillstheir own version, first seen Oct 25, 2026
- manhquydev/flow-skilltheir own version, first seen Oct 16, 2026
- es6kr/skillstheir own version, first seen Oct 13, 2026
- manuelcozar55/setup-claude-codetheir own version, first seen Oct 13, 2026
- ririversoza/claude-harnesstheir own version, first seen Oct 11, 2026
Used with
Projects with it also use retro, verify and 3 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
AWSSkillretro
Conduct a retrospective on a coding session.
Also in 3 projects that use this
Microsoft AzureSkillverify
End-to-end app testing pipeline for browser-based web apps on cloud browsers via Azure Playwright Workspaces: verify the live app, author tests for working flows, heal failing tests, then re-run with video and produce one consolidated report. Three phases (app verification, test authoring, test healing) plus a scoped final sanity run; each phase emits a sub-report. Use when asked to "verify my app", "make sure my app works and has good test coverage", "do a full QA pass", "add tests where I'm missing coverage", "fix failing tests and add new ones", "smoke test before deploy", "QA my deployment", or any combined verify-and-test ask. Requires Azure CLI (`az login`) and a Playwright Workspace resource ID — bootstraps configs on first invocation. Browser-based web apps with Playwright only — not for unit/API tests, native mobile/desktop, load/perf, or full accessibility audits.
Also in 3 projects that use this
GoogleSkillcode-review
Adversarial systems, concurrency, memory-model, and architecture code review protocol. Use this skill whenever asked to review code, PRs, diffs, or branches in Filament.
Also in 4 projects that use this
- GoogleSkill
grill-me
A relentless interview to sharpen a plan or design.
Also in 3 projects that use this
SentrySkillskill-creator
Create new agent skills following the Agent Skills specification. Use when asked to "create a skill", "add a new skill", "write a skill", "make a skill", "build a skill", or scaffold a new skill with SKILL.md. Guides through requirements, writing, registration, and verification.
Also in 3 projects that use this
Other versions35 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 35
- revfactory/harness★ 9.1ksame name
- cexll/myclaude★ 2.8ksame name
- Simpleyyt/ai-manus★ 1.6ksame name
- BankrBot/openclaw-skills★ 1.2ksame name
- BankrBot/skills★ 1.2ksame name
- bam-bam-2/solo-skills★ 366same name
- serejaris/personal-corp-os★ 243same name
- serejaris/ris-claude-code★ 243same name
- SensorsIot/Embedded-AI-Harness★ 235same name
- skillsynchq/txcript★ 208same name
25 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.
0
Common
Most of the 24 edited versions we read keep these.
1
Distinctive
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 79
Never bypass permissions on a real repo.
- Very specific: line 100
No product map, no features.md / backlog.md, no campaign ledger
“Common” is measured: kept by at least 6 in 10 of the 24 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Harness
One method: seven stations. Humans pick a door; the agent walks the stations underneath and stops at every gate a human owns.
Humans read USAGE.md. Agents read this file plus references/invariants.md, then only the files the current step names.
Invocation
Explicit only. Never auto-start.
| Tool | Invoke |
| Cursor, Claude Code | /harness |
| Codex | @harness |
| Human says | Route | File |
Bare /harness, first time | Setup only, then ask the pile | playbooks/bootstrap.md |
Bare /harness, later | Ask the pile. Do not redo Setup | doors.md |
/harness bug … | Bug lane | playbooks/bug.md |
/harness feature … | Feature lane | playbooks/feature.md |
/harness campaign … | Campaign | playbooks/campaign.md |
"yes", "go on", /harness continue | Next station | playbooks/continue.md |
/harness loop / loop stop | Arm or disarm the stop hook | playbooks/loop.md |
A station name (/harness diagnose) | That station only | table below |
Bare /harness never codes and never starts Campaign. Talk to the human per
human-handback.md: lead with the question
they can answer, use station names, not file names.
Seven stations
Load only the playbook for the current station. Each owns one job and returns one artifact.
| # | Station | Playbook | Returns | Must not |
| 1 | Setup | bootstrap.md | Status + skill index in AGENTS.md or CLAUDE.md, resume in .harness/ | Map the product, code |
| 2 | Investigate | diagnose.md | Report with cites, empty diff | Pick destination, code |
| 3 | Orient | landscape.md | Destination evidence | Implement |
| 4 | Plan | propose.md → contract.md | Accepted five-part contract | Edit |
| 5 | Build | implement.md | Diff inside the contract | Widen scope |
| 6 | Prove | verify.md → review.md | Evidence + review; merge stays human | Next feature |
| 7 | Harden | standardize.md → scale.md | One durable layer, or handoff | New campaign |
Templates: status-block.md, investigate-checklist.md, task-contract.md, tracker.md, resume.md.
Standing rules
Full list in invariants.md. Short form:
- Preflight before writing any instruction file.
- Five-part contract before any behavior change. Empty diff until then.
- One station per turn. Bug and feature lanes may chain to the next human gate; each station still returns its artifact.
- Cite files, commands, results. Never invent.
- Tag each move Delegate / Review / Own. Humans own destination, architecture, security, compatibility, merge.
- End every hand-back with a reviewer note and a trust close that names a mechanism.
- Never bypass permissions on a real repo. Stop at secrets.
- Skill index points at installed skills; never copy packs in.
Routing
Specialists: routing.md. Surface notes: surfaces/. Method notes for teaching: ladders.md, dro.md, control-layers.md, case-studies.md.
Completion criteria
First /harness (Setup) ends with:
- Preflight evidence: folder, clean diff or named dirt, one check result
- Status + skill index shown before write, then written to one file
.harness/resume.mdwritten,.harness/gitignored- Install verified or gap named
- Reviewer note + trust close
- The question: bug, one feature, or a big piece?
- No product map, no
features.md/backlog.md, no campaign ledger
Later turns end with the current station's artifact, the updated resume, the next human question, reviewer note, trust close.
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.