repo-governance
Skill10 starsChanged 32 days agoMITView on GitHub
Use when repository complexity, delivery, ownership, security, release, or incident signals change; when governance may be missing; or when adding, reviewing, deferring, or upgrading repo-seed capabilities
In this repository6 other files it gives its agents.
- AGENTS.mdAGENTS.md
- docs/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- repo-decisionsSkill
- repo-reviewSkill
- repo-seedSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 3 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Run node scripts/audit-governance.mjs --json and read .repo-seed/manifest.json.
- Very specific: line 14
Check whether a recommendation's observed facts and safety boundary actually apply to the current change.
- Distinctive: line 34
Check whether an existing or external mechanism already satisfies the need.
- Very specific: line 36
Ask for authorization. Record enabled, external, deferred, or declined plus the assessment hash and concise reason.
- Distinctive: line 38
Update review instructions when the capability adds a new risk surface.
- Distinctive: line 49
Do not convert every preference into a gate.
- Very specific: line 53
When .repo-seed/manifest.json is absent, a read-only audit and adoption proposal are allowed.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Evolving repository governance
Governance grows with the repository. Do not install every mechanism up front, and do not wait for a preventable failure before naming a new control.
Start with a read-only assessment
Run node scripts/audit-governance.mjs --json and read .repo-seed/manifest.json. The audit observes discrete repository facts only; it does not authorize changes.
- Check whether a recommendation's observed facts and safety boundary actually apply to the current change. A
blockinglabel alone does not make an optional capability a task-wide gate; a confirmed safety dependency blocks only the affected action. - An
advisoryrecommendation improves maturity without making the current work unsafe. Finish the requested work, then raise it in the handoff. - A suppressed recommendation was already declined or deferred against the same facts. Do not ask again until its assessment hash changes.
Missing optional tools or unavailable external evidence do not block independent authorized work. Reuse valid evidence, bound retries, and report the exact unresolved dependency. Do not repeat approval already given for the same scope.
Authority boundary
You may autonomously inspect, audit, compare capability state, and produce a dry-run. Ask the user before you:
- enable a capability or add its files;
- install or replace a Git hook;
- change governance policy or a source of truth;
- connect an external system or change remote permissions.
Once a capability is enabled, untouched repo-seed-owned files may refresh through the normal upgrade channel. Preserve user-owned content and stop on conflicts.
Midstream capability workflow
- Name the new repository facts and the capability they make relevant.
- Check whether an existing or external mechanism already satisfies the need.
- Explain benefit, ongoing cost, files/processes added, and whether the recommendation blocks current work.
- Ask for authorization. Record
enabled,external,deferred, ordeclinedplus the assessment hash and concise reason. - Preview with a dry-run, apply only the approved capability, then run
node scripts/run-gates.mjs. - Update review instructions when the capability adds a new risk surface.
Promotion ladder
Promote knowledge to the cheapest layer that reliably prevents recurrence:
- one task-specific fact stays in its spec or plan;
- a repeated project convention becomes resident instruction or a focused skill;
- a rule that must hold without exception becomes a deterministic gate;
- a systemic escaped failure becomes a postmortem linked to the new guardrail.
Do not convert every preference into a gate. Deterministic enforcement is for objective rules with low false-positive cost.
Unmanaged repositories
When .repo-seed/manifest.json is absent, a read-only audit and adoption proposal are allowed. Never write the Core automatically. The adoption sequence is audit, dry-run, user authorization, incremental write, then verification. Preserve an existing source of truth rather than creating a duplicate.
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.

