agent-flow
AGENTS.md8 starsChanged 17 days agoMITView on GitHub
This file is for Codex reviewing changes to the agent-flow plugin itself. Claude Code sessions in this repo use .claude/CLAUDE.md instead. The general co-review rubric — your role next to Lawliet, the output contract (one verdict line + SEVERITY: file:line: issue findings), the severity scale, what to defer to Lawliet, and the tie-breaker — ships with the plugin at templates/codex/review-rubric.md and is inlined into every review prompt by scripts/dispatch-codex-review.sh. Follow it. The checklist below adds the blocker classes specific to…
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 repository12 other files it gives its agents.
- .claude/CLAUDE.mdCLAUDE.md
- agent-behavior-constraintsSkill
- agentsview-usageSkill
- evidence-driven-workSkill
- explainer-design-systemSkill
- exploration-strategySkill
- graphify-usageSkill
- personal-kb-usageSkill
- prompt-refinementSkill
- skill-agent-mappingSkill
- task-classificationSkill
- verification-gatesSkill
The file
We split this AGENTS.md into the rules it gives an agent
1 rule. About 1 is common, 0 are distinctive, and 0 are very specific to this project.
≈1
Common · estimate
Many files like this have a rule of this kind.
≈0
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
- could be automatic: a setting could enforce it
- Common: line 31
No secrets: no API keys, tokens, passwords, or credentials in committed files.
≈ “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 co-review — agent-flow repo checklist
This file is for Codex reviewing changes to the agent-flow plugin itself.
Claude Code sessions in this repo use .claude/CLAUDE.md instead.
The general co-review rubric — your role next to Lawliet, the output
contract (one verdict line + SEVERITY: file:line: issue findings), the
severity scale, what to defer to Lawliet, and the tie-breaker — ships with
the plugin at templates/codex/review-rubric.md and is inlined into every
review prompt by scripts/dispatch-codex-review.sh. Follow it. The checklist
below adds the blocker classes specific to this repository.
Project context
Agent Flow is a multi-agent orchestrator plugin for Claude Code. Agents are
markdown files under agents/; orchestration commands live in commands/;
skills under skills/; the Phases 3–5 loop is the plugin workflow
workflows/implement-review-verify.js. /orchestrate delegates each phase
to a specialist agent (Riko, Senku, Loid, Lawliet, Alphonse).
Repo-specific blocker classes
- Shell safety: standalone
.shfiles must useset -euo pipefailat the top (hook scripts that intentionally fail open may useset -uo pipefailwith a comment). Flag a new.shfile missing it. ERROR. Embedded Bash insidecommands/*.mdis out of scope. - Heredoc variable expansion:
commands/*.mdBash blocks must not use$VARinside<<'PROMPT'heredocs — single-quoted delimiters suppress expansion, so the variable is emitted literally. ERROR. - YAML validity: YAML emitted to
.claude/orchestration.local.mdmust be syntactically valid; unclosed keys, bad indentation, or stray characters break the grep-based parsers. ERROR. - No hardcoded paths: scripts must not contain
/Users/...or other machine-specific absolute paths — use${HOME},$(git rev-parse --show-toplevel), or${CLAUDE_PLUGIN_ROOT}. WARNING. - No secrets: no API keys, tokens, passwords, or credentials in committed files. ERROR.
- Hook output contracts: PreToolUse denials use
hookSpecificOutput.permissionDecision: "deny", never top-levelcontinue: false(which halts the whole session); Stop hooks emit{"decision": "block", ...}only when blocking and stay silent otherwise (sole exception: the{systemMessage}bypass notice when.claude/skip-test-verificationexists); a hook's stdout must be exactly one JSON object (tool output goes to stderr). ERROR. - Workflow script rules:
workflows/*.jsmust keepexport const metaa pure literal as the first statement, must not useDate.now(),Math.random(), arglessnew Date(), orimport(), and behavior changes need a matching scenario inscripts/test-implement-review-verify.js. ERROR for the forbidden APIs, WARNING for a missing test.
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.

