did-ai-really-fix-it
AGENTS.md0 starsChanged 50 days agoMITView on GitHub
This repo is a Claude Code plugin: skills/verify-agent-claims/SKILL.md is the skill, skills/verify-agent-claims/scripts/verify_claims.py its script, and .github/workflows/verify-claims.yml an optional GitHub Action. It's public, and one of three Cairn plugins whose shared principles and evidence are in nicuk/cairn-principles. Each rule names what enforces it, or says that nothing does. The audit warns on a rule here that does neither. Everything in .github/workflows/self-test.yml runs on every push. - Every check the script gains gets a planted case in --self-test, and is broken once…
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 32,671 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 repositoryOne other file it gives its agents.
- verify-agent-claimsSkill
The file
We split this AGENTS.md into the rules it gives an agent
3 rules. About 0 are common, 0 are distinctive, and 3 are 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.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 27
The skill's frontmatter must parse as strict YAML: Never put : inside the description.
- Very specific: line 32
README
##sections are shared across the family: don't add, rename or reorder one here alone. - Very specific: line 37
Nothing private goes in this repo: no client or product names, no private codebases, no absolute paths from anyone's machine.
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 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
Cairn Verify: rules for agents working here
This repo is a Claude Code plugin: skills/verify-agent-claims/SKILL.md is the skill,
skills/verify-agent-claims/scripts/verify_claims.py its script, and
.github/workflows/verify-claims.yml an optional GitHub Action. It's public, and one of
three Cairn plugins whose shared principles and evidence are in nicuk/cairn-principles.
Each rule names what enforces it, or says that nothing does. The audit warns on a rule here
that does neither. Everything in .github/workflows/self-test.yml runs on every push.
Rules
- Every check the script gains gets a planted case in
--self-test, and is broken once on purpose to prove the case fails. Enforced byskills/verify-agent-claims/scripts/verify_claims.py: its self-test fails when a check on its list stops firing. That a new check joins the list is not enforced: add it in the same commit as the check. - The script is read-only and can't reach the network.
--smokeruns the user's own command in a temporary worktree that is always removed; it kills only its own child by PID. No networking import, in any form, is enforced by.github/scripts/check_privacy.py. The worktree cleanup is enforced by the self-test'ssmokecase. That the script writes nothing else, and thatPRIVACY.mdstill describes it, is not enforced: re-read both whenever either changes. - The skill's frontmatter must parse as strict YAML. Never put
:inside the description. Enforced by.github/scripts/check_frontmatter.py, becauseclaude plugin validatedoesn't. - Numbers the docs state about the repo come from their source. Enforced by
.github/scripts/check_claims.py: every "N checks" in the README,plugin.jsonand the demo images must equal the count the self-test reports. - README
##sections are shared across the family: don't add, rename or reorder one here alone. Enforced by the daily drift check innicuk/cairn-principles. - Every release raises
versionin.claude-plugin/plugin.json, with an entry inCHANGELOG.mdand a git tag. Version and changelog are enforced by.github/scripts/check_claims.py. The tag is not enforced: tagvX.Y.Zwhen the release merges. - Nothing private goes in this repo: no client or product names, no private codebases, no
absolute paths from anyone's machine. Absolute paths are enforced by
.github/scripts/check_privacy.py. Names and private code are not enforced: check before committing. - This repo passes Cairn Memory's audit with no warnings. Enforced by the last step of
.github/workflows/self-test.yml, pinned to a Memory release. Raise the pin on purpose, once this repo passes the newer release.
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.

