flow
Skill9 starsChanged 28 days agoMITView on GitHub
Run the buildflow gated build process from idea to real done-evidence. Walk gated stages (Idea->Research->Scope->PRD->ADR->Contract->Cards->Build->Review->Deploy/Ship->Verify->Retro), each with a honest gate that must pass before advancing. Adapts to project type (web|cli|library|skill). Use when starting or driving a real product build, when the user types /flow, /flow next, /flow card, /flow check, or asks to scope/plan/ship a project through gates. Kill at any gate is a valid outcome.
33 other projects use a version of this.
Adopted by
Most recently taken up by
- phuthuycoding/kanban-flowtheir own version, first seen Nov 8, 2026
- baodhg/FLOWtheir own version, first seen Nov 8, 2026
- jocarrd/crux-agentstheir own version, first seen Nov 8, 2026
- cofin/flowtheir own version, first seen Nov 7, 2026
- retospect/precis-mcptheir own version, first seen Nov 6, 2026
- luckys/agent-skillstheir own version, first seen Nov 1, 2026
Used with
Projects with it also use grill, implement and 5 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
renatorozasSkillgrill
Adversarial code review. Don't let me ship until the changes pass your scrutiny.
Also in 3 projects that use this
Hugging FaceSkillimplement
Make tests pass. Invoke after /write-tests produces failing tests.
Also in 5 projects that use this
SentrySkillreview
Fixture review skill.
Also in 6 projects that use this
AnthropicSkillstatus
Case status summary by audience — client-facing (plain language), internal (for the professor), or court-ready (formal caption format per local rules). Same facts, different framing and depth. Use when a student needs to update the client, brief the professor, or prepare a court status report.
Also in 3 projects that use this
BlockSkillplan
Draft a TDD, PRD, or lightweight plan in the conversation without committing it.
Also in 3 projects that use this
AWSSkillship
Finalize dev work on an issue branch — update docs, push, and open a PR.
Also in 3 projects that use this
Other versions33 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 33
- react/react★ 252ksame name
- facebook/react★ 252ksame name
- software-mansion/react-native-audio-api★ 856same name
- simota/agent-skills★ 96same name
- panayiotism/claude-harness★ 79same name
- rainday/smart-blog-skills★ 35same name
- cofin/flow★ 19same name
- basezh/agent-skills-on-base★ 15same name
- retospect/precis-mcp★ 11same name
- foryouself83/harness-tier★ 10same name
23 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository3 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/flow/law/CLAUDE.mdCLAUDE.md
- harnessSkill
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 0 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 26 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 52
Never mark a card done without pasted world-state evidence.
- Very specific: line 53
Never edit _templates/ or runner/flow.sh during a project run.
“Common” is measured: kept by at least 6 in 10 of the 26 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
/flow — buildflow gated build harness
Idea to world-state done-evidence (deployed URL for web; install+run for cli/library/skill), not paperwork. Each stage has an artifact and a GATE. Kill at any gate is a valid, honored outcome.
Idea -> Research -> Scope -> PRD -> ADR -> Contract -> Cards -> Build -> Review -> Deploy -> Verify-live -> Retro
|------------------ planning (files in flow/) ------------------| |------- shipping (inside cards/) -------|Two-layer harness
/flow is two layers. Both must agree; the script can pass hollow content.
- Mechanical —
runner/flow.sh(deterministic, exit 0/1). Stage/card lifecycle plus cheatable checks: unchecked boxes (including leftover- [ ]under## Open decisions),[FILL], card status, empty done-evidence, plus content scanners on 01/02/03/05 (typed-web entry floor, L-above-A, PRD adjectives, OD cap). Always run it first — its exit code is ground truth, never host judgment. Open-decision leftover boxes are the same scanner;clarifyonly prints them — not a second gate. - Semantic — this skill. The script cannot tell a real competitor quote from a
fabricated one, or a grade-laundered C feature from an honest B. After mechanical
PASS, apply
references/gate-examples.mdbefore letting the operator advance.
PTC_ONLY
This file is always-on. Load one extra reference for the verb in play (load table)
plus that verb's row in references/command-dispatch.md. Do not load the rest until
needed. Typed verbs always win over chat.
STOP
- Runner reports BLOCKED by another session's lock → STOP. Never
FLOW_FORCEa live session; concurrent runs corrupt the plan. SetFLOW_SESSION_IDper session. - Mechanical FAIL → STOP. Relay exact line numbers. In
teach, never tick a box or write an artifact on the operator's behalf. - Security-class skip/debt → HALT until the operator accepts the exposure in
DEBT.md. - Never mark a card
donewithout pasted world-state evidence. - Never edit
_templates/orrunner/flow.shduring a project run. NEXT_VERB=/ JSONnext_verbis advisory; never auto-execnext,card,card-start,check,fix-gate,auto, orskip.
Run flow.sh first
From the project root (where flow/ and cards/ live):
# macOS / Linux / Windows Git Bash:
bash <skill-dir>/runner/flow.sh <command>
# Windows PowerShell or cmd (INCLUDING inside Codex): use the .cmd launcher, NOT bare bash.
<skill-dir>\runner\flow.cmd <command>Bare bash on Windows/Codex is usually WSL and cannot read C:/ paths. Always
runner/flow.cmd from a Windows shell; it locates Git Bash. Call flow.sh directly
only when bash is Git Bash.
Flow v1 runtime requirement (pre-release target): product operations require Python
3.7 or newer (standard library only), resolved as python3 then python. Missing,
below-floor, or nonworking interpreters refuse progression; a nonempty
FLOW_HARNESS_DISABLE also refuses product operations. Help and doctor remain available
for diagnosis. The graph executor owns ordering; the mechanical runner remains the gate
verdict authority. This contract is not shipped until implementation, migration/rollback,
review and CI acceptance are complete; see the v1 migration guide.
<skill-dir> is the install home (~/.claude/skills/flow, ~/.codex/skills/flow,
~/.agents/skills/flow, Antigravity ~/.gemini/antigravity-cli/skills/flow (CLI) /
~/.gemini/config/skills/flow (IDE), or project .claude/skills/flow). Override with
FLOW_PROJECT_ROOT. Antigravity (agy CLI / IDE): same SKILL.md bundle; run
agy inspect to confirm it loaded.
One session per project. flow/.lock auto-reclaims after FLOW_LOCK_TTL (default
900s). Mutating commands refuse a fresh foreign lock. Export a stable FLOW_SESSION_ID.
FLOW_FORCE=1 only for a lock known dead; unlock clears it.
Three laws
- Inspect first. Competitors, live systems, code. Evidence, not vibes.
- Contract is the seam. Stage 05 before any code. Backend builds TO it, UI consumes FROM it. Amend the contract, then code. Honor a shape now (null/stub) even when the value ships later.
- Done = proof in the world. Name type-aware done-evidence up front: deployed URL for web; install+run for cli/library/skill. "Tests pass" / "code merged" are mid-pipeline, never done.
Load table
| In play | Load |
| typed verb | references/command-dispatch.md (that row only) |
| plain language | references/concierge.md, references/flow-catalog.tsv |
| next PASS / check semantic | references/gate-examples.md |
| card / build session | law/CLAUDE.md |
| UI | law/DESIGN.md |
| auto | references/auto-run.md |
| attest / receipts | references/attestations.md |
| mode work | references/mode-work.md |
| eval | references/gate-eval.md |
| Idea/Scope ritual | references/forge-idea.md |
| engines | references/codex-integration.md, references/antigravity-integration.md, references/claudekit-skills.md |
| parallel occupancy | references/host-agnostic-parallel.md |
| agents | references/agent-detection.md, references/agent-stage-mapping.md |
| skip / halt | references/debt-and-halts.md |
| host context adapters (opt-in) | references/host-hooks.md |
| harness | harness/README.md |
| retro | law/RETRO.md |
Dispatch rules
- Entering a project mid-cycle? Fresh session and existing
flow/orcards/→ run resume first (read-only session-story: last session, in-flight + dwell, gate state, oneNEXT ->line). Skip only with live context this session. - Always call
flow.shfirst and relay exit + output. Lock BLOCKED → STOP. - On next: FAIL → stop, offer help, never author/tick in teach. PASS → semantic
challenge in
gate-examples.md(the one extra file). Do not silently advance hollow content; do not silently block a sound artifact. - On card/check:
law/CLAUDE.md— one card per session, only## Allowed files, contract is the seam, done = world-state proof. Run recall first; apply its output. - Mode
work: interview once, draft 00–05, pause only for scope sign-off, same gates as teach. - Never edit
_templates/orrunner/flow.shduring a project run. Read any file the runner just created before editing it.
Seams (one-line pointers)
- Verb map:
references/command-dispatch.md(no Commands table in this file). - Chat front door:
references/concierge.md+references/flow-catalog.tsv. - Idea/Scope persona ritual:
references/forge-idea.md(adapted from BMAD-METHOD,bmad-forge-idea, MIT, opt-in, never a gate condition). - Codex second engine:
references/codex-integration.md. - Antigravity Gemini-3 third engine:
references/antigravity-integration.md(confirm load withagy inspect). - ck/ak skill layer:
references/claudekit-skills.md. - Parallel occupancy:
references/host-agnostic-parallel.md. - Receipts:
references/attestations.md. harness/is flow-owned. Improve via R-IMPROVE-HARNESS inreferences/native-rituals.md. Neverstory update --status implemented— usestory complete --proof-source.
Forbidden
- Checking a gate box or writing a planning artifact on the operator's behalf (teach).
- Setting a card
donewithout pasted world-state evidence. - Building two cards in one session, or in parallel before
readymarks them safe. - Frontend code before the UI mock card is approved.
- Editing
_templates/orrunner/flow.shduring a project run.
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.

