autoform-bot
AGENTS.md136 starsChanged 19 days agoMITView on GitHub
User instructions override this file. CONTRIBUTING.md covers contribution, style, and validation; the development skill covers product-development workflow and cross-file maintenance. - Name the user outcome or maintenance burden before fixing, restacking, or merging a proposal; a request to resolve review blockers does not justify it. - Optimize for repository quality and a simple product, not PR count, green checks, or docs and tests that only validate their own planning state. - Closing or replacing a proposal that removes duplication is…
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 repository7 other files it gives its agents.
- agent-reviewSkill
- debrief-orchestratorSkill
- develop-pluginSkill
- formalizeSkill
- human-reviewSkill
- roadmapSkill
- setupSkill
The file
We split this AGENTS.md into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 20
Treat current main as the product baseline unless an explicitly documented stack says otherwise.
≈ “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
AutoformBot agent guidance
User instructions override this file. CONTRIBUTING.md covers contribution, style, and validation; the development skill covers product-development workflow and cross-file maintenance.
Product judgment
- Name the user outcome or maintenance burden before fixing, restacking, or merging a proposal; a request to resolve review blockers does not justify it.
- Optimize for repository quality and a simple product, not PR count, green checks, or docs and tests that only validate their own planning state.
- Closing or replacing a proposal that removes duplication is a success. Keep reusable insight in the owning skill or an issue; remove an artifact only when authorized, and otherwise recommend removing it.
Repository invariants
- Treat current
mainas the product baseline unless an explicitly documented stack says otherwise. Unmerged replacements do not supersede it. - Develop the plugin here and exercise installed behavior in an independent Lean consumer. The bundled Cabannes thesis is never a source of product exceptions.
- Markdown under
blueprint/roadmap/is the only roadmap and dependency graph. Implementation handoffs live only in per-article files underblueprint/.implementation-notes/; rendered sites, dashboards, and runtime projections are derived views (CLI reference). - In consumer formalization work, Lean imports are allowed; Formalize owns that policy. Never replace it with a blanket import ban.
- Public CLI and JSON shapes are contracts. Lean server work requires an explicit project root, bounded time and output, verified descendant cleanup, and no retry after dispatch with an unknown outcome. A project root is not an OS sandbox.
- Claims are optional and coordinate through Git
refs/autoform-claims/*; the dashboard's publication overlay is loopback-only, ephemeral, and kept out of the vault and public site. Pushing claims or publishing requires user-supplied authority. - The bundled example mirrors only the artifacts listed under generated mirrors; ignore rules are not among them.
- Note each change worth knowing about under Unreleased in
CHANGELOG.md, and keep release versions synchronized acrosspyproject.toml,uv.lock, Claude and Muse manifests, and the semantic base of Codex's cachebuster version.
Inform, don't break
Autoform's commands inform the orchestrator; they must not break a codebase (#325). A new check that a repository could fail warns and keeps the exit status, so a repository that passed before still passes and the orchestrator decides. A command exits nonzero only when it cannot give a correct answer: unreadable input, a stale or missing build it would otherwise misreport, a contract it cannot honor, or evidence it would have to forge. Never add a refusal to the graph loader, where it stops every command. The CLI reference gives each command's exit statuses and warnings.
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.

