G
Gabriel-Valin/harness-bootstrap
Agent Skill that makes any git repo agent-ready: verification layers, sealed per-task state, handoffs and an independent evaluator.
★ 2·updated 2026-11-16
View on GitHub ↗Link checked 2026-11-17 · HTTP 200 · opens the vendor's own site
After install: how to tell it works
- What you see after install — The command exits 0 and the terminal shows no "command not found" or "permission denied".
- One-line self-check — Run the smallest usage example from that repository's README. If the output matches what the README describes, it is working.
- If that does not work — This project has no troubleshooting notes. Search its issues for "install".
The install command comes from the repository; these checks describe observable results only and do not replace the project's own docs.
- Star count
- 2 ★ · Gabriel-Valin/harness-bootstrap · read from the GitHub search API · (checked 2026-11-19 20:00:48)
- Source
- read directly from its GitHub page · type: open-source project
- Harvest history
- first seen 2026-11-17 · last seen 2026-11-20 · seen 4× · listed by 2 sources
- Score
- score 203 · 2 ★ · 2 sources · repo
- Install
- see Install & config above
FAQ (answers come from our records)
Who is it for?
We filed it under these tags: Agent skills. The tags come from the entry itself, not a taxonomy we imposed.
Who maintains it?
repository
Gabriel-Valin/harness-bootstrap;first catalogued 2026-11-17;last seen 2026-11-20。There is no official certification or quality rating here — we record sources and dates only.README excerpt (English, untranslated) · Read the full README on GitHub →
# harness-bootstrap
Excerpted from the repository README (abridged)
An Agent Skill that turns any git repository into one where coding
agents can work across sessions, worktrees and people without forgetting where they stopped, and
without calling something "done" before a script proves it.
It works with any agent that supports Agent Skills: Claude Code, Codex, Cursor, GitHub Copilot,
Gemini CLI, OpenCode, Pi, Devin, Windsurf, Kiro and others.
## Install
bash
npx skills add Gabriel-Valin/harness-bootstrap -g
The skills CLI detects your agents and installs the skill
where each one reads it. -g installs it for your user, which fits a skill you run once per
repository; leave it out to install into the current project only. -a picks agents, and
@ pins a version (npx skills add Gabriel-Valin/harness-bootstrap@v1.0.0 -g).
Then, in a repository, ask your agent to set up the agent harness (or invoke
/harness-bootstrap in agents with slash commands).
## What it installs in your repository
The skill reads the repo first, asks only what it cannot find out (for example which agents your
team uses), and never overwrites a file you already have.
| Layer | Where | Committed |
| Entry file | AGENTS.md, plus pointers such as CLAUDE.md for agents that need one | yes |
| Docs | docs/ARCHITECTURE.md, PRODUCT.md, RELIABILITY.md, docs/decisions/ | yes |
| Contract | harness/: commands per verification layer, architecture rules, checklist, rubric | yes |
| Verification | scripts/harness/: init.sh, verify.sh, feature.py, check_architecture.py, cleanup-scan.sh | yes |
| Lifecycle skills | .agents/skills/harness-{start,handoff,cleanup,evaluate}, mirrored for agents that read another folder | yes |
| Agent adapters | Optional extras, for example a subagent and hooks for Claude Code | yes |
| Task state | .harness/tasks/ /: feature list, contract, progress, handoff, run logs | no |
Day to day, your agent uses four skills that ship inside the repository, so every teammate gets
them on clone:
harness-start: agree on scope, turn the done criteria into items with exact verification
commands, and record that they fail before any code is written.
harness-handoff: verify, clean up and write a handoff that the next session (or teammate)
can resume from with nothing but the repo.
harness-evaluate: an independent, skeptical check, run in a subagent or a fresh session,
before anything is called done.
harness-cleanup: periodic maintenance that turns repeated review feedback into executable
checks.
## Guarantees that do not depend on your agent
An item becomes passing only by running its verification command through feature.py. The
task state is sealed on every save, so a hand edit makes every command refuse to work until
feature.py recover sends the passing items back to be verified again. Evidence logs are
hashed.
Verification that runs zero tests or skips tests counts as a failure, not a pass.
Agent-specific hooks (Claude Code today) only catch hand edits earlier. The harness works
agents can work across sessions, worktrees and people without forgetting where they stopped, and
without calling something "done" before a script proves it.
It works with any agent that supports Agent Skills: Claude Code, Codex, Cursor, GitHub Copilot,
Gemini CLI, OpenCode, Pi, Devin, Windsurf, Kiro and others.
## Install
bash
npx skills add Gabriel-Valin/harness-bootstrap -g
The skills CLI detects your agents and installs the skill
where each one reads it. -g installs it for your user, which fits a skill you run once per
repository; leave it out to install into the current project only. -a picks agents, and
@ pins a version (npx skills add Gabriel-Valin/harness-bootstrap@v1.0.0 -g).
Then, in a repository, ask your agent to set up the agent harness (or invoke
/harness-bootstrap in agents with slash commands).
## What it installs in your repository
The skill reads the repo first, asks only what it cannot find out (for example which agents your
team uses), and never overwrites a file you already have.
| Layer | Where | Committed |
| Entry file | AGENTS.md, plus pointers such as CLAUDE.md for agents that need one | yes |
| Docs | docs/ARCHITECTURE.md, PRODUCT.md, RELIABILITY.md, docs/decisions/ | yes |
| Contract | harness/: commands per verification layer, architecture rules, checklist, rubric | yes |
| Verification | scripts/harness/: init.sh, verify.sh, feature.py, check_architecture.py, cleanup-scan.sh | yes |
| Lifecycle skills | .agents/skills/harness-{start,handoff,cleanup,evaluate}, mirrored for agents that read another folder | yes |
| Agent adapters | Optional extras, for example a subagent and hooks for Claude Code | yes |
| Task state | .harness/tasks/ /: feature list, contract, progress, handoff, run logs | no |
Day to day, your agent uses four skills that ship inside the repository, so every teammate gets
them on clone:
harness-start: agree on scope, turn the done criteria into items with exact verification
commands, and record that they fail before any code is written.
harness-handoff: verify, clean up and write a handoff that the next session (or teammate)
can resume from with nothing but the repo.
harness-evaluate: an independent, skeptical check, run in a subagent or a fresh session,
before anything is called done.
harness-cleanup: periodic maintenance that turns repeated review feedback into executable
checks.
## Guarantees that do not depend on your agent
An item becomes passing only by running its verification command through feature.py. The
task state is sealed on every save, so a hand edit makes every command refuse to work until
feature.py recover sends the passing items back to be verified again. Evidence logs are
hashed.
Verification that runs zero tests or skips tests counts as a failure, not a pass.
Agent-specific hooks (Claude Code today) only catch hand edits earlier. The harness works
Listed badge (optional)
Copy this line into the README if you want the listing mark. Optional, and it does not affect being listed.
[](https://anclum.com/en/s/gabriel-valin-harness-bootstrap/)
- Source
- github · https://github.com/Gabriel-Valin/harness-bootstrap
- Browse by field
- Agent skills
- Traction
- 2 ★
- Listed by
- 2 independent sources
- added
- today
More entries
Filed under
Every tag opens the full list for that filter. Counts are the live catalogue figures.