repo-seed
AGENTS.md10 starsChanged 32 days agoMITView on GitHub
repo-seed: a cross-tool skill that seeds agent-native, self-governing repositories This repository's governance layer is managed by repo-seed; product code remains repository-owned. .repo-seed/manifest.json is authoritative for managed files, capability state, governance paths, external sources of truth, and the installed repo-seed version. Route governance work explicitly: If automatic skill activation is unavailable, read the linked resident SKILL.md directly and follow it. See docs/architecture.md for the module map and seams. Directories named there are the canonical layout; anything else is user code. Write…
- Reads credentials
- Commits and pushes
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 33,315 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository6 other files it gives its agents.
- docs/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- repo-decisionsSkill
- repo-governanceSkill
- repo-reviewSkill
- repo-seedSkill
The file
We split this AGENTS.md into the rules it gives an agent
8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 17
If automatic skill activation is unavailable, read the linked resident SKILL.md directly and follow it.
- Distinctive: line 33
Run relevant tests while working and the shared gate runner before every commit.
- Distinctive: line 37
Run the gates before every commit.
- Distinctive: line 46
Never git commit or git push unless the user explicitly asks.
- Very specific: line 47
Never modify files outside the seeded paths (AGENTS.md, CLAUDE.md, docs/, scripts/, .agents/skills/, .github/, CONTRIBUTING.md, LICENSE, …
- Very specific: line 48
Never read .env files or other secrets.
- Distinctive: line 52
Follow docs/AGENTS.md: one home per fact, tutorials vs references, hygiene checklist.
- Distinctive: line 64
Follow docs/testing.md. Name the regression risk, then choose the smallest test set at the lowest sufficiently real boundary.
≈ “Common” is an estimate: at least 1 in 5 of the 42,283 AGENTS.md 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
AGENTS.md
repo-seed: a cross-tool skill that seeds agent-native, self-governing repositories
Repository governance
This repository's governance layer is managed by repo-seed; product code remains repository-owned. .repo-seed/manifest.json is authoritative for managed files, capability state, governance paths, external sources of truth, and the installed repo-seed version.
Route governance work explicitly:
- Change or pull-request review must use
repo-review. - A durable choice with meaningful alternatives must use
repo-decisions. - New complexity, delivery, ownership, security, release, or incident signals must use
repo-governance. - Seeding, adoption, governance-layer upgrades, or upstream repair must use the global
repo-seedskill. If it is unavailable, stop and ask rather than hand-copying upstream governance. - Ordinary implementation follows this file and the linked project documents without invoking repo-seed again.
If automatic skill activation is unavailable, read the linked resident SKILL.md directly and follow it.
Repository layout
See docs/architecture.md for the module map and seams. Directories named there are the canonical layout; anything else is user code.
Code readability
Write code for human readers. Keep the primary logic path visible; introduce an abstraction only when it hides more complexity than its indirection adds; choose names that reveal domain meaning, effects, and cost; use one repository term per concept; and keep behavior within its owning boundary.
Commands
- Test:
node --test scripts/*.test.mjs - Lint:
npm run lint (if configured) - Gates:
node scripts/run-gates.mjs
Run relevant tests while working and the shared gate runner before every commit. CI or an installed pre-commit hook may repeat the exhaustive matrix.
Governance loop (hard rules)
- Run the gates before every commit; install the optional pre-commit capability only with user authorization.
- Record a decision in docs/decisions/ when a change chooses among meaningful alternatives whose rationale may be revisited; decisions are not a change log.
- Risk-boundary changes start from an Approved spec in docs/specs/; routine changes are exempt.
- A subtle, systemic, or costly escaped failure earns a postmortem in docs/postmortems/ linked to a permanent guardrail.
- Implementation materially derived from an external source retains provenance at the closest stable repository location; follow docs/development.md.
- The only upgrade channel for the seeded governance layer is re-running the repo-seed skill. Never hand-edit seeded files to "match upstream"; re-run the skill instead.
Security rules
- Never
git commitorgit pushunless the user explicitly asks. - Never modify files outside the seeded paths (AGENTS.md, CLAUDE.md, docs/, scripts/, .agents/skills/, .github/, CONTRIBUTING.md, LICENSE, .editorconfig, .gitattributes, .repo-seed/) without asking.
- Never read
.envfiles or other secrets.
Documentation
Follow docs/AGENTS.md: one home per fact, tutorials vs references, hygiene checklist.
Decisions
Durable architecture and process decisions use MADR in docs/decisions/. Status flows Proposed → Accepted → Superseded by NNNN. A superseded record is never rewritten into its opposite.
Governance evolution
Use the repo-governance skill when complexity, delivery, ownership, security, release, or incident signals change. Read-only audits and dry-runs are autonomous. Ask before enabling a capability, installing a hook, changing policy/source-of-truth, or connecting an external system. Raise blocking recommendations before implementation and advisory recommendations in the handoff; do not repeat a declined recommendation until its assessment hash changes.
Testing
Follow docs/testing.md. Name the regression risk, then choose the smallest test set at the lowest sufficiently real boundary; verify observable outcomes rather than implementation details.
Skills
.agents/skills/repo-review— semantic review policy (instantiated per project) before merging..agents/skills/repo-decisions— how to write and update decision records..agents/skills/repo-governance— progressive capability assessment, authorization, and upgrades.
Enabled optional capabilities
- CI runs the gates on every push/PR: .github/workflows/ci.yml.
- AI-assisted commits/PRs disclose participation per docs/ai-disclosure.md.
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.

