Agents
AGENTS.md1 starsChanged 58 days agoApache-2.0View on GitHub
Prose the repo guards. Some of it is immutable, some is generated, none of it is a scratch pad. This directory holds the charter, the decision record, plans, runbooks and two generated artifacts. README.md is the index of what lives where; STYLE.md is the taxonomy. This file is the set of rules that turn a harmless documentation edit into a failing build. - ADRs are immutable. A decision that changed course is a new ADR that supersedes the old one,…
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 31,774 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 repository39 other files it gives its agents.
- agent-core/AGENTS.mdAGENTS.md
- .agents/AGENTS.mdAGENTS.md
- AGENTS.mdAGENTS.md
- behavioral-regression/AGENTS.mdAGENTS.md
- claude-foundation/AGENTS.mdAGENTS.md
- config/AGENTS.mdAGENTS.md
- corpora/AGENTS.mdAGENTS.md
- demo/AGENTS.mdAGENTS.md
- examples/AGENTS.mdAGENTS.md
- experiments/AGENTS.mdAGENTS.md
- flow-corpus/AGENTS.mdAGENTS.md
- flow-protocol/AGENTS.mdAGENTS.md
- openspec/AGENTS.mdAGENTS.md
- scripts/AGENTS.mdAGENTS.md
- skills/AGENTS.mdAGENTS.md
- tests/AGENTS.mdAGENTS.md
- claude-foundation/CLAUDE.mdCLAUDE.md
- merge-gate-auditorSkill
- update-executive-reportSkill
- update-executive-reportSkill
- architecture-drift-guardSkill
- commonSkill
- corpus-guardianSkill
- dataset-lintSkill
- deploySkill
- e2e-matrix-sentinelSkill
- eval-corpus-forgeSkill
- hierarchical-recursive-brainstormSkill
- model-benchSkill
- openai-judgeSkill
- openspec-implementation-reviewSkill
- openspec-peer-reviewSkill
- openspec-quality-planSkill
- pre-pr-gateSkill
- project-setupSkill
- quality-gateSkill
- refactoring-decomposerSkill
- repo-invariant-reviewSkill
- test-completeness-guardSkill
The file
We split this AGENTS.md into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
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
- Very specific: line 70
An unlisted plan is an invisible plan: Every plans/<topic>/ must appear in README.md's plans table.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 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 — docs
Prose the repo guards. Some of it is immutable, some is generated, none of it is a scratch pad.
This directory holds the charter, the decision record, plans, runbooks and two generated
artifacts. README.md is the index of what lives where;
STYLE.md is the taxonomy. This file is the set of rules that turn a harmless
documentation edit into a failing build.
Map
| Path | Role |
STYLE.md | The taxonomy: which kind of doc a change is, and where it goes |
CHARTER.md | North-star scope and invariants. Links and claims are both gated |
decisions/ | Immutable ADRs. Numbering is not contiguous |
plans/ | plans/<topic>/{PLAN.md,REVIEW.md}; each must be listed in README.md |
runbooks/ | Operational how-to prose |
e2e-matrix/, matrix-coverage.md | GENERATED. Regenerate, never hand-edit |
golden-corpus/ | Human-label contract; stays empty until a human writes rows |
Diagram
flowchart TD
accTitle: routing a documentation change to the right file
accDescr: A decision becomes a new immutable ADR, a how-to becomes a runbook, a plan of work becomes a plan folder that must be indexed in the docs README, and an overview becomes a component README.
Q["a change worth writing down"]
subgraph D["docs/ (you are here)"]
ADR["decisions/NNNN-*.md<br/>immutable"]
RB["runbooks/"]
PL["plans/<br/>one folder per topic"]
IDX["README.md index"]
GENA["matrix-coverage.md<br/>e2e-matrix/<br/>GENERATED"]
end
RM["a component README"]
SUP["a NEW superseding ADR"]
Q --> ADR
Q --> RB
Q --> PL
Q --> RM
PL --> IDX
RB --> IDX
ADR --> SUP
GENA --> IDX
classDef here fill:#e8f0fe,stroke:#1a73e8,stroke-width:2px
class D hereRules that bite here
- ADRs are immutable. A decision that changed course is a new ADR that supersedes the old one, never an edit to it. Editing an accepted ADR destroys the record of why the old choice was made, which is the only thing an ADR is for.
- ADR numbers are not contiguous, and
0007is an intentional gap. Do not backfill it and do not renumber later ADRs to close it. Take the next free number after the highest one indecisions/README.md, and add your row to that index. CHARTER.mdis doubly gated.scripts/check_charter_drift.pyfails when its markdown links rot;scripts/check_charter_invariants.pymechanically re-checks its claims — package roles, invariants, default-off flags — against the code. A charter edit that states something the repo does not do fails CI, not review.e2e-matrix/andmatrix-coverage.mdare generated and freshness-gated. Regenerate withpython tests/test_matrix_coverage.py --updateandmake e2e-matrix-update. A hand edit is erased and failsmake matrix-check/make e2e-matrix-checkin the meantime.- An unlisted plan is an invisible plan. Every
plans/<topic>/must appear inREADME.md's plans table; eleven of them went unreachable before that index existed. The same applies to a new runbook.
Verify
make invariantsSubagents
| Task in this directory | Agent | Why |
| Find every doc that cites an ADR before superseding it | explorer | Read-only Grep across prose; supersession leaves stale citations behind |
| Run the charter and freshness gates after a docs change | test-runner | Has Bash; make invariants chains three checkers and only one will fail |
| Review a charter or ADR edit before pushing | narrow-critic | An overstated invariant is a correctness problem no linter reads |
See also
| Doc | Read it when |
STYLE.md | You are creating a doc and must decide whether it is an ADR, runbook, plan or README |
decisions/README.md | You need the next free ADR number or the supersession convention |
README.md | You added a plan or runbook and have to make it reachable |
CHARTER.md | A change might exceed the declared scope or touch an invariant and needs escalation |
plans/agents-md-directory-docs/TEMPLATE.md | You are writing or editing an AGENTS.md anywhere in this repo |
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.

