Agent-Workflow-Kit
AGENTS.md1 starsChanged 4 months agoMITView on GitHub
docs/ai is the source of truth for rules, skills, commands, agent roles, and artifacts. Before working, read: - .agent-workflow-kit/config.conf — project settings, active profiles, model roles, check command, and base ref. - docs/ai/README.md — map of the AI surface. - docs/ai/rules/project.md — tool-agnostic invariants. - docs/ai/rules/coding-rules.md — this stack's conventions. - docs/ai/rules/commit-message.md — Conventional Commits. <check> means CHECK_COMMAND from the config. Run it after substantive changes unless it is explicitly none. Do not commit or push without an explicit request.
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 31,700 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 repository11 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- .cursor/rules/main.mdcCursor rule
- skill-ai-auditSkill
- skill-buildSkill
- skill-choose-modelSkill
- skill-deps-auditSkill
- skill-learnSkill
- skill-planSkill
- skill-review-correctnessSkill
- skill-review-designSkill
- skill-review-securitySkill
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 37
If the user writes /review, treat it as a possible native Codex command, not a project workflow.
- Very specific: line 55
Write plans to docs/ai/plans/YYYY-MM-DD-short-task.plan.md.
- Very specific: line 56
Write reviews to docs/ai/reviews/YYYY-MM-DD-short-task.review.md.
≈ “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
Codex Adapter
docs/ai is the source of truth for rules, skills, commands, agent roles, and
artifacts. Before working, read:
.agent-workflow-kit/config.conf— project settings, active profiles, model roles, check command, and base ref.docs/ai/README.md— map of the AI surface.docs/ai/rules/project.md— tool-agnostic invariants.docs/ai/rules/coding-rules.md— this stack's conventions.docs/ai/rules/commit-message.md— Conventional Commits.
<check> means CHECK_COMMAND from the config. Run it after substantive changes
unless it is explicitly none. Do not commit or push without an explicit request.
What Codex picks up
AGENTS.md— project instructions..agents/skills/*/SKILL.md— repo-scoped skills..codex/agents/*.toml— repo-scoped custom subagents..codex/config.toml— base Codex settings (sandbox, network).
Text triggers
Codex does not use the project slash commands in .claude/commands. When the user
writes a text trigger or invokes a skill via $..., run the matching workflow:
| Trigger | Workflow |
codex plan, $skill-plan | docs/ai/commands/plan.md |
codex review, $skill-review-correctness | docs/ai/commands/review-full.md |
codex choose-model, $skill-choose-model | docs/ai/commands/choose-model.md |
codex ai-audit, $skill-ai-audit | docs/ai/commands/ai-audit.md |
codex deps-audit, $skill-deps-audit | docs/ai/commands/deps-audit.md |
codex learn, $skill-learn | docs/ai/commands/learn.md |
If the user writes /review, treat it as a possible native Codex command, not a
project workflow. The project full review is orchestrated by a supported host via
review-full;
Codex performs the correctness/security passes inside it, per
docs/ai/skills/review-correctness.md and docs/ai/skills/review-security.md.
Codex roles
- Planning:
docs/ai/agents/codex-plan.md. - Correctness review:
docs/ai/agents/review-correctness.md. - Security review:
docs/ai/agents/review-security.md. - Custom subagents:
.codex/agents/codex-plan.toml,.codex/agents/review-correctness.toml,.codex/agents/review-security.toml,.codex/agents/review-design.toml. - Canonical CLI invocations:
docs/ai/agents/cli.md— do not invent flags.
Artifacts
- Write plans to
docs/ai/plans/YYYY-MM-DD-short-task.plan.md. - Write reviews to
docs/ai/reviews/YYYY-MM-DD-short-task.review.md. - Return a short summary and the file path to chat.
Model policy
Resolve model roles and effort through .agent-workflow-kit/config.conf and
docs/ai/commands/choose-model.md. Do not pin model IDs in this adapter.
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.