agentleFS
lautioliverlautioliver/barrilete-kit/.claude/skills/plan-feature/SKILL.md

plan-feature

Skill6 starsChanged 7 days agoMITView on GitHub

Turns an idea or big feature into a researched technical plan and then into an epic with issues and dependencies. It writes no code. Use with "/plan-feature", "plan X", "I want to do X, how do we approach it?" or before any work that does not fit in a single PR.

31 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use explore, review and 3 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.
  • genomewalkerSkill

    explore

    RLM-style recursive memory exploration - dynamically navigate the memory graph

    Also in 3 projects that use this

  • SentrySkill

    review

    Fixture review skill.

    Also in 5 projects that use this

  • AWSSkill

    ship

    Finalize dev work on an issue branch — update docs, push, and open a PR.

    Also in 3 projects that use this

  • OpenAISkill

    systematic-debugging

    Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes

    Also in 3 projects that use this

  • GoogleSkill

    code-review

    Adversarial systems, concurrency, memory-model, and architecture code review protocol. Use this skill whenever asked to review code, PRs, diffs, or branches in Filament.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 30 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before proposing anything, read:

    line 12
  2. Distinctive:

    If the idea is vague on something that changes the design (who uses it, what happens in the edge cases), ask now, before writing the plan.

    line 19
  3. Distinctive:

    Write it with these sections.

    line 23
  4. Distinctive:

    No issue goes out with undefined values ("a reasonable window", "an adequate limit"): either the plan fixes them, or the issue carries an …

    line 60

“Common” is measured: kept by at least 6 in 10 of the 30 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Plan a feature

It avoids starting to code too early. First you understand what is there, then you decide, and only at the end do you split into issues. This skill touches no code.

1. Research

Before proposing anything, read:

  • docs/development/architecture.md, docs/reference/ (domain, modules, surfaces) and the related ADRs.
  • The glossary (docs/reference/glossary.md) and the contexts map (docs/reference/contexts.md): the plan uses their terms and names the contexts it touches. The rules are in docs/decisions/kit/ADR-003-minimal-ddd-glossary-and-contexts.md.
  • The code of the modules that will be touched: how it is built today, not how it should be.
  • docs/development/roadmap.md and horizon.md: was it already planned? Does it clash with anything?
  • Related open issues (gh issue list --search …) and .claude/skills/project-status/available.sh.

If the idea is vague on something that changes the design (who uses it, what happens in the edge cases), ask now, before writing the plan.

2. Technical plan

Write it with these sections. Be concrete: real files and modules, not generic ones.

## Goal
What is achieved and for whom. What is explicitly left out.

## Current state
How what will be touched works today (with paths).

## Affected components
Modules, tables, endpoints, screens. Mark the sensitive ones (docs/agents-map.json). Name the bounded contexts they touch (docs/reference/contexts.md).

## Decisions needed
What has to be decided before building, with options and a recommendation.
The hard-to-revert ones → propose an ADR. A new dependency between contexts (not in `depends_on`) is one of them.

## Risks
What can go wrong (data, money, security, performance, active users) and how it is mitigated.

## Migrations
Schema changes and their strategy (skill db-migration). "None" if there are none.

## Tests
What has to be tested and at what level.

## Issues
The split into PRs, in order, with their blockers.

Show the plan and stop. The decisions in the "Decisions needed" section are taken by the user or the team, not the agent. Iterate until they approve it.

3. Split into issues

With the plan approved:

  • Each issue = one PR that is reviewed on its own and leaves the system working. Typically 3 to 8.
  • Blockers only where there really is an order (schema → API → screen). Parallel things, with no blocker.
  • If there are pending decisions or an ADR to write, they are the first issue and are loaded as a blocker (blocked_by) of the issues that depend on that decision. Having the epic say so in its text is not enough: prepare.sh only sees loaded blockers.
  • No issue goes out with undefined values ("a reasonable window", "an adequate limit"): either the plan fixes them, or the issue carries an Open decisions section with the concrete question for the team.
  • Destructive migrations in two steps (expand / contract) are two issues.

4. Create

The full technical plan goes in the epic's body: that way it stays next to the issues and /implement-issue reads it when taking each one. Do not create a separate document in docs/ unless it is an ADR. It is created with planner.py (in this folder; requires python3):

  1. Write the plan in a temporary JSON (outside the repo):
    {
      "epic": {"title": "Migrate accounts to the new table", "body": "## Goal\n…", "labels": ["type:feature", "area:api"]},
      "issues": [
        {"key": "schema", "title": "Create the accounts table", "body": "…", "labels": ["type:task", "area:infra"]},
        {"key": "api", "title": "Expose accounts in the API", "body": "…", "labels": ["type:task", "area:api"],
         "blocked_by": ["schema"], "reason": "needs the table"},
        {"key": "screen", "title": "Show accounts in the dashboard", "body": "…", "labels": ["type:feature"],
         "blocked_by": ["api", 25], "reason": "consumes the endpoint; #25 defines the design"}
      ]
    }
    blocked_by accepts plan keys or numbers of existing issues. Titles, labels and bodies follow the convention of the create-issue skill (steps 3–6).
  2. Validate and show the draft:
    .claude/skills/plan-feature/planner.py <plan.json> --draft
    It validates labels against labels.yml, exactly one type: per issue, existing keys and numbers, and that there are no cycles. It shows the issues in the order they can be done. If it fails, fix the plan; do not skip the validation. Show the user that output and ask for an explicit "yes".
  3. Create:
    .claude/skills/plan-feature/planner.py <plan.json>
    It creates the epic, each issue in order, links them as sub-issues, records the blockers, adds status:blocked and the Blocked by #N line where it applies. If it fails halfway, it reports what it managed to create: show it and do not retry the whole plan (it would duplicate issues); create only what is missing.

5. Deliver

Link to the epic, the issues in order and what can be started now (.claude/skills/project-status/available.sh). The next step is /implement-issue <n> on the first available one.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.