agentleFS
MannyFlussMannyFluss/claude-skills/decompose-issue/SKILL.md

decompose-issue

Skill0 starsChanged 7 months agoMITView on GitHub

Break a large GitHub issue into atomic ralph-sized sub-issues (one behaviour, one session each). Use when an issue is too big to implement in one go, when implement-issue flags it as oversized, or when you want to split an epic into a story queue.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    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 a split, understand what the issue actually touches.

    line 20
  2. Distinctive:

    No hidden dependency on another sub-issue being done first (or explicitly sequenced if there is one)

    line 31
  3. Distinctive:

    Ask the user: "Does this split look right?

    line 44
  4. Distinctive:

    Only after the user approves the split.

    line 48
  5. Distinctive:

    After all children are created, edit the parent issue body to append a progress checklist.

    line 61
  6. Distinctive:

    If the parent was just a planning placeholder with no real body, replace the body entirely with the checklist.

    line 76
  7. Distinctive:

    Do not close the parent — leave it open so it tracks overall progress as children close.

    line 78

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Decompose Issue

Break one fat issue into a queue of atomic stories, each small enough for a single /implement-issue session.

Step 1 — Read the issue

gh issue view <NUMBER> --json title,body

Also read progress.txt if it exists — prior context may clarify scope.

Step 2 — Explore the codebase for scope

Before proposing a split, understand what the issue actually touches. Read relevant files, check existing tests, look at related issues:

gh issue list --label ralph --state open --json number,title

Step 3 — Propose the split

Present a numbered list of proposed sub-issues. Each one must have:

  • A single, unambiguous acceptance condition ("done" is obvious)
  • Changes confined to one area of the codebase
  • No hidden dependency on another sub-issue being done first (or explicitly sequenced if there is one)

Format:

Proposed split of #N — <original title>:

1. <sub-title> — <one-line acceptance condition>
2. <sub-title> — <one-line acceptance condition>
3. <sub-title> — <one-line acceptance condition>

Suggested order: 1 → 2 → 3 (reason: <dependency or risk rationale>)

Ask the user: "Does this split look right? Any changes before I create the issues?"

Step 4 — Create sub-issues on confirmation

Only after the user approves the split. Create each child with a consistent header so the tree is traceable:

gh issue create \
  --title "<sub-title>" \
  --body "> Sub-issue of #<PARENT>\n\n## Acceptance criteria\n<condition>\n\n## Context\n<anything the implementer needs to know>" \
  --label ralph

Capture each new issue number as it's created — you need them for Step 5.

Step 5 — Update the parent with a task checklist

After all children are created, edit the parent issue body to append a progress checklist. This makes the tree visible from the top and GitHub renders it as a progress bar:

# Get current body first
gh issue view <PARENT> --json body

# Then edit — append the checklist to the existing body
gh issue edit <PARENT> --body "<existing body>

## Sub-issues
- [ ] #A — <sub-title>
- [ ] #B — <sub-title>
- [ ] #C — <sub-title>"

If the parent was just a planning placeholder with no real body, replace the body entirely with the checklist. If it's a real deliverable, append after the existing content.

Do not close the parent — leave it open so it tracks overall progress as children close.

Step 6 — Hand off

Tell the user:

Decomposed #N into: #A — "<title>", #B — "<title>", #C — "<title>"

Start with: Read progress.txt then implement issue #A

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.