agentsclimarketplace

Workflow composer

Skill neon-straw/brief-spec-overlays/skills/workflow-composer

Compose a single non-overlapping workflow across custom planning skills, mattpocock/skills, and Superpowers. Use when a user asks which skills to run, wants to combine brief-writer and to-spec with matt-pocock-skills or superpowers, is moving from a vague idea toward implementation, or risks creating duplicate briefs, PRDs, specs, plans, or issues.From its SKILL.md

Install
npx -y skills add neon-straw/brief-spec-overlays --skill workflow-composer

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

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 0 stars0 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.

SKILL.md

6.0 KB, ~1.3k tokens by cl100k_base, as published. Nobody here has run it

Workflow Composer

Use this skill to choose the next skill sequence, not to create another planning document. The goal is one clear path from idea to code, with one source of truth at each stage.

First Move

  1. Read the user's request and any named skill invocation.
  2. Inspect repo guidance if available: AGENTS.md or CLAUDE.md, CONTEXT.md or CONTEXT-MAP.md, relevant ADRs, configured issue-tracker instructions, and existing planning docs.
  3. Identify the current stage:
    • vague idea
    • scoped brief
    • PRD-ready feature
    • complex spec-needed feature
    • issue-ready work
    • implementation/debugging/review
  4. Recommend or run the smallest next skill sequence. Do not run two skills that produce competing artifacts for the same stage.

Skill Roles

Use the custom skills for early product shape:

  • brief-writer: turn vague intent into a compact brief and scope boundary.
  • to-spec: turn resolved PRD/brief context into a feature-local spec pack when the feature is too complex for PRD-only execution.
  • workflow-composer: choose the sequence and prevent duplicate planning layers.

Use mattpocock/skills for engineering workflow:

  • setup-matt-pocock-skills: configure issue tracker, triage labels, and domain-doc layout.
  • grill-with-docs: stress-test a plan against domain language and update CONTEXT.md or ADRs.
  • grill-me: stress-test a plan without updating docs.
  • prototype: validate risky state, API, or UI assumptions with throwaway code.
  • to-prd: create a PRD from already-discussed context.
  • to-issues: split a PRD/spec/plan into vertical implementation issues.
  • tdd: implement one issue with red-green-refactor.
  • diagnose: debug a bug or regression through a disciplined reproduction loop.
  • triage: classify incoming issues and make them ready for agents.
  • zoom-out / improve-codebase-architecture: understand or improve architecture after real pressure appears.

Use Superpowers when the user explicitly wants that methodology, when the repo already tracks Superpowers artifacts, or when its execution/review discipline is the clearest fit:

  • brainstorming: explore unclear creative/product direction.
  • writing-plans: produce an implementation plan from accepted requirements or specs.
  • executing-plans: execute an approved plan with checkpoints.
  • test-driven-development: use Superpowers' TDD loop if that is the chosen local standard.
  • verification-before-completion: verify before claiming completion.
  • requesting-code-review / receiving-code-review: review and respond to review feedback.

Routing

For a vague idea:

brief-writer -> optional grill-with-docs -> to-prd -> optional to-spec -> to-issues -> tdd

For a vague idea where the user explicitly asks for Superpowers:

brief-writer -> superpowers:brainstorming -> superpowers:writing-plans -> superpowers:executing-plans

For a feature already discussed in the chat:

to-prd -> optional to-spec -> to-issues -> tdd

For a complex feature with multiple modules, external APIs, state machines, or contracts:

to-prd -> to-spec -> to-issues

Tell to-issues to read PRD.md, spec/, and plan/ if to-spec created them.

For a concrete implementation issue:

tdd

For a bug, failing test, performance regression, or strange behavior:

diagnose

For repo confusion or architecture orientation:

zoom-out -> optional improve-codebase-architecture

For a small concrete edit with clear requirements:

no planning skill -> implement -> verification-before-completion if available

Source of Truth Ladder

Keep one canonical artifact per layer. Resolve tracker-specific paths from repo guidance first: AGENTS.md or CLAUDE.md, the issue-tracker configuration provided by setup-matt-pocock-skills, GitHub/GitLab issues, and the active branch may override the local examples below.

LayerCanonical artifact
Scope and whydocs/brief.md or the feature brief
Domain languageCONTEXT.md and docs/adr/
Product requirementsthe configured PRD location, for example .scratch/<feature>/PRD.md or a linked GitHub issue/PRD
Complex design/contractsthe configured spec location, for example .scratch/<feature>/spec/ or docs linked from the tracking issue
Execution phasesthe accepted plan: local plan/, Superpowers plan, GitHub issue checklist, or PR plan
Work itemsthe configured tracker: GitHub issues, local .scratch/<feature>/issues/, or another repo-defined tracker
Active implementation statethe active branch, PR branch, worktree, code, tests, and verified behavior

If two artifacts claim the same role, stop and choose the newer or explicitly accepted one. If implementation contradicts a spec or ADR, update the durable artifact or record the drift.

Anti-Duplication Rules

  • Do not create both a Superpowers plan and a separate phase plan unless one explicitly supersedes the other.
  • Do not run to-prd if a current accepted PRD already exists; update it only if the user asks.
  • Do not run to-spec for simple CRUD, copy edits, small bug fixes, or one-file changes.
  • Do not use brainstorming after scope is accepted unless new uncertainty appears.
  • Do not let to-issues split from PRD alone when spec/ and plan/ exist.
  • Do not turn every decision into an ADR. Promote only durable, surprising, or hard-to-reverse decisions.

Response Shape

When advising the user, answer with:

  1. Current stage.
  2. Next skill sequence.
  3. Source-of-truth artifact to read or create.
  4. The first concrete command/invocation or action.

Keep it short. This skill exists to remove workflow fog, not to add another ceremony layer.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 326,750. 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.