Spec-Flow
AGENTS.md94 starsChanged 6 months agoMITView on GitHub
Spec-Flow coordinates multiple coding tools through a single shared canon. Every agent must read the same rules, reuse the same artifacts, and only write in its designated areas. - .spec-flow/ holds the canonical workflow docs, repo map, state schemas, memory, and automation scripts. - In consumer projects, epics/<slug>/state.yaml is the single source of truth for epic phase progress; feature-level specs/<feature>/state.yaml is optional and must mirror the epic. - Roadmaps, specs, plans, tasks, and ship notes all inherit the phase order:…
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 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.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 26
Never fork new “top-level religions.” Add new integrations under their dedicated folder and mirror the shared canon from .spec-flow/.
- Very specific: line 27
All tools may read .claude/ for reference, yet non-Claude edits are prohibited.
- Very specific: line 31
When progressing any phase, update the relevant state.yaml and stop at the documented review boundaries (no macro “run everything” flows).
- Distinctive: line 34
Follow these rules and every agent—Claude, Codex, Cursor, or future additions—will operate from the same source of truth.
≈ “Common” is an estimate: at least 1 in 5 of the 36,368 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
Spec-Flow Agents
Spec-Flow coordinates multiple coding tools through a single shared canon. Every agent must read the same rules, reuse the same artifacts, and only write in its designated areas.
Shared brain
.spec-flow/holds the canonical workflow docs, repo map, state schemas, memory, and automation scripts.- In consumer projects,
epics/<slug>/state.yamlis the single source of truth for epic phase progress; feature-levelspecs/<feature>/state.yamlis optional and must mirror the epic. - Roadmaps, specs, plans, tasks, and ship notes all inherit the phase order:
spec → clarify → plan → tasks → implement → optimize → preview → ship → finalize.
Tool boundaries
.claude/— Claude Code-only prompts, commands, and hooks. Readable by other tools, but only Claude edits this tree..codex/— Codex CLI-specific prompts, commands, adapters, and skills. Codex writes here; other tools leave it alone.- Additional tool adapters must be labeled explicitly until their source-repo surfaces exist.
- Shared assets may live elsewhere in active repo surfaces such as
docs/,bin/,.github/, and.spec-flow/, but every tool must honor.spec-flow/repo-map.yamland.spec-flow/domains/*.yamlwhen deciding where to add code.
Hard rules
- Never fork new “top-level religions.” Add new integrations under their dedicated folder and mirror the shared canon from
.spec-flow/. - All tools may read
.claude/**for reference, yet non-Claude edits are prohibited. - In consumer projects, epic work begins under
epics/<slug>/; features live inspecs/<feature>/and must link back to their parent epic insidestate.yaml. - When progressing any phase, update the relevant
state.yamland stop at the documented review boundaries (no macro “run everything” flows). - Persistent learnings or roadmap shifts belong in
.spec-flow/memory/and should be reflected in the shared docs before modifying tool-specific trees.
Follow these rules and every agent—Claude, Codex, Cursor, or future additions—will operate from the same source of truth.
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 28,693 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 repository48 other files it gives its agents.
- .codex/Agents.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- analysis-phaseSkill
- anti-duplicationSkill
- audit-claude-mdSkill
- breaking-change-detectorSkill
- caching-strategySkill
- clarification-phaseSkill
- create-agent-skillsSkill
- create-hooksSkill
- create-meta-promptsSkill
- create-slash-commandsSkill
- create-subagentsSkill
- debug-phaseSkill
- dependency-conflict-resolverSkill
- design-tokensSkill
- domain-memorySkill
- dry-runSkill
- epic-meta-promptingSkill
- epic-scopingSkill
- epic-sprintsSkill
- epic-walkthroughSkill
- error-recoverySkill
- finalize-phaseSkill
- git-workflow-enforcerSkill
- hallucination-detectorSkill
- implementation-phaseSkill
- mockup-extractionSkill
- multi-agent-votingSkill
- optimization-phaseSkill
- parallel-execution-optimizerSkill
- planning-phaseSkill
- production-deployment-phaseSkill
- project-initialization-phaseSkill
- regression-test-generatorSkill
- roadmap-integrationSkill
- session-boundarySkill
- shadcn-integrationSkill
- specification-phaseSkill
- staging-deployment-phaseSkill
- staging-validation-phaseSkill
- task-breakdown-phaseSkill
- task-status-syncSkill
- theme-consistencySkill
- ultrathinkSkill
- workflow-detectionSkill
- worktree-contextSkill
- x-announcementSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
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.