gmickel/gmickel-claude-marketplace
Overview
This skill creates structured epics and tasks in .flow/ from a feature request, Flow ID, or short description. It is a plan-first tool: it produces specs, tasks, and acceptance criteria but does not write implementation code. Trigger it with /flow-next:plan plus text or Flow IDs (fn-1-abc, fn-1-abc.2, or legacy fn-1).
How this skill works
The planner parses your arguments, optional flags (research, review, depth), and the configured review backend, then runs a guided question flow when needed. It uses the bundled flowctl to read/write all plan artifacts under .flow/ and enforces repo conventions, task sizing, and non-implementation rules. Research modes (repo-scout or context-scout) and review modes (codex, rp, export, none) are supported and influence depth and follow-up actions.
When to use it
- Turning a feature idea or bug report into an epic and discrete tasks
- Refining an existing Flow epic (fn-N or fn-N-xxx) or a specific task (fn-N-xxx.M)
- Preparing work for Ralph autonomous or multi-model review gates
- Before implementation work to avoid drift and ensure consistent gating
- When you need repo-aware plans that integrate with flowctl and .flow/
Best practices
- Always use the bundled flowctl wrapper: FLOWCTL="${CLAUDE_PLUGIN_ROOT}/scripts/flowctl" $FLOWCTL <cmd>
- Keep each task sized for one /flow-next:work iteration (~100k tokens max); split large efforts
- Do not include implementation code in plans—use signatures, patterns, and references only
- Store all artifacts only under .flow/ (.flow/epics, .flow/tasks, .flow/specs); never use external TODO trackers
- Choose research and review options deliberately (defaults: standard depth, repo-scout, review backend if configured)
Example use cases
- /flow-next:plan Add OAuth login for users — produce epic, tasks, acceptance criteria, rollout notes
- /flow-next:plan fn-1-abc — expand legacy epic into tasks and validate sizing
- /flow-next:plan fn-2.3 then review via /flow-next:plan-review — refine a single task and queue review
- /flow-next:plan --research=rp --review=codex --depth=deep New multi-tenant config model
- /flow-next:plan fn-5 then review with rp — prepare exportable spec for RepoPrompt gates
FAQ
The skill warns to run /flow-next:setup to refresh local scripts but continues planning; this check is informational only.
Can the planner write code or implementation details?
No. Plans may include function signatures, file references, patterns, and non-obvious gotchas, but must not contain implementation code.
13 skills
This skill helps you convert feature requests or Flow IDs into structured epic plans and task breakdowns using flowctl
This skill coordinates Carmack-level plan reviews using rp-cli context building and chat to streamline architecture and design assessments.
This skill helps manage git worktrees and copy environment files to enable parallel feature work, isolated reviews, and clean, reproducible workspaces.
This skill helps you set up flowctl locally and document CLAUDE.md instructions to enable reliable flow-next workflows.
This skill executes a plan file or Beads epic end-to-end with structured setup, tests, and commit workflows to ensure completion.
This skill assesses codebase readiness across eight pillars, reports production gaps, and fixes agent-readiness issues to accelerate reliable AI development.
This skill helps you explore code efficiently with RepoPrompt CLI, reducing tokens while locating auth patterns, data flows, and API endpoints.
This skill automates browser tasks using agent-browser to open pages, snapshot accessibility refs, interact with elements, and verify UI across flows.
This skill performs Carmack-level plan reviews of Flow epic specs and design docs using rp or codex backends to ensure production-ready plans.
This skill conducts a thorough interview to extract complete implementation details from a spec or bead and refines the document.
This skill helps you manage git worktrees efficiently and copy .env files across parallel feature branches for isolated review.
This skill scaffolds or updates a repo-local Ralph harness to enable opt-in autonomous coding workflows.
This skill helps you trigger manual plan-sync to refresh downstream task specs after implementation drift.