agentsclimarketplace
Sign in

Flow

Skill cofin/flow/skills/flow

Use when a repository has .agents, when the user asks for Flow lifecycle routing, Beads-backed task memory, spec-first planning, TDD implementation, sync/status, review, finish, archive, or /flow:* help.From its SKILL.md

Install
npx -y skills add cofin/flow --skill flow

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 15 stars15 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
  • runs commandsInstructs the agent to run 2 commands, including `make validate-skills` and 1 more.

SKILL.md

3.0 KB, 691 tokens by cl100k_base, as published. Nobody here has run it

Flow Router

Flow coordinates Context-Driven Development in .agents/ repositories. Keep this skill small: use it to identify the active lifecycle phase, enforce the Beads-first invariants, and load the matching lifecycle skill.

Flow is a skill, not a CLI. There is no flow executable. Never run flow, flow sync, flow prd, etc. as shell commands. Invoke this skill (or the matching lifecycle skill), or use the /flow:* slash commands where the harness supports them.

Beads mode: Skip every bd invocation when the SessionStart hook reports Beads Backend: Missing (None) or Disabled via plugin config (useBeads=false). Treat spec.md markers as fallback source of truth and skip /flow:sync. Never halt for missing Beads. See references/discipline.md.

Workflow

  1. Check hook-provided Flow context first; otherwise detect .agents/, Beads (bd), git branch, and repo-native commands.
  2. Route the request:
    • Setup, validation, install, context initialization: use flow-setup.
    • PRD, research, plan, refine, revise, task creation: use flow-planning.
    • Implement, claim ready tasks, TDD, commit, task close: use flow-execution.
    • Sync, status, refresh, cleanup, context drift: use flow-sync-status.
    • Review, finish, archive, revert, docs, phase completion: use flow-completion.
  3. Record durable discoveries and task state in Beads. Markdown files are synchronized views.
  4. Prefer repo-native commands from .agents/workflow.md or hook context for validation.

Guardrails

  • Never edit task markers ([ ], [~], [x], [!], [-]) manually in spec.md.
  • Do not run bd export, auto-stage, bd dolt push, or git operations through Beads unless .agents/beads.json allows it or the user asks.
  • Store Flow specs and planning artifacts under .agents/specs/<flow_id>/.
  • Make minimal targeted changes and record findings with bd note <id> "..." when work exceeds a quick fix.

Validation

  • For planning: verify the plan is decision-complete before presenting it.
  • For implementation: verify red-green-refactor evidence, full relevant tests, and Beads task closure before claiming completion.
  • For sync/status: read backend state first and report drift instead of guessing.
  • For this repository: run make validate-skills and make validate-codex-manifest after skill or command changes.

References Index

Example

User: "Use Flow to implement the current spec."

Action: load flow-execution, claim a ready Beads task, add investigation notes, follow TDD, close the task with evidence, then sync according to policy.

What ships with it: 20 files

101.1 KB alongside SKILL.md

agents/

references/

Gives 0 of the 12 instructions most plan spec skills give in 691 tokens

Counted across 1,360 of the 2,617 authors here whose files we hold, read 2026-09-06

  • Ask one question at a timein 73 of 1360
  • Write the spec using the templatein 22 of 1360
  • Ask clarifying questions if neededin 19 of 1360, across 18 files
  • Wait for user confirmation before proceedingin 19 of 1360
  • Save plans to the plans directoryin 17 of 1360, across 13 files
  • Check for product marketing context firstin 16 of 1360, across 5 files
  • Read the plan file completelyin 16 of 1360
  • Order tasks by dependencyin 16 of 1360
  • Gather context from the conversationin 15 of 1360, across 9 files
  • Explore the codebase instead of askingin 15 of 1360, across 13 files
  • Wait for explicit user approvalin 14 of 1360, across 13 files
  • Quiz the user on the breakdownin 13 of 1360, across 7 files

Said here and by no other author read

  • Route operations to owning lifecycle skills
  • Load only the selected lifecycle skill
  • Preserve Git history and tag immutability

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.