Large File Refactor

This skill is coordination-exempt: skip the ai-coord gate for its declared work.

If a slash or dollar invocation already added these instructions to the conversation, follow them directly. In that case, do not invoke this skill again through a skill tool.

Use LOC thresholds to discover candidates, then decide whether a split is justified by cohesion, coupling, ownership, and change risk. Source files match above 1000 LOC. Test files use a relaxed 2000 LOC discovery threshold.

Arguments

• path: Optional file or directory to scan. Default: current working directory.
• --include-generated: Include generated, vendored, dependency, and build-output paths that are skipped by default.

Workflow

1. Resolve the skill directory, then run the helper from the target repository:

   uv run "<skill-dir>/scripts/large-file-refactor.py" [path] [--include-generated]
   

   Copy

2. Preserve the helper's Markdown table as the exhaustive report. Do not omit matching rows, even when the refactor plan only covers a subset.

3. If the helper reports no threshold matches, stop after the report. A match is a candidate, not proof that the file should be split.

4. Draft a refactor plan for the 3 largest files only, unless the user explicitly requested another count.

5. For each candidate, rank split value by mixed responsibilities, change frequency/risk, coupling, and testability. Use whichever semantic symbol/reference tooling is available. When Serena is installed, prefer it:

   • Inspect symbol overviews, references, imports, and relevant history.
   • Use the evidence to choose extraction boundaries, target module names, migration order, and test coverage.

6. This workflow defaults to a report and plan. Implement only when the user's request already authorizes execution or the user subsequently approves it. Do not require a separate request for an unchanged, already authorized outcome.

Refactor Plan Format

For each selected file, include:

• Current role: the file's apparent responsibility and why line count is a symptom.
• Semantic pass: the exact symbol/reference/history inspection to run before moving code.
• Split proposal: 2-5 target modules or files with responsibilities.
• Migration order: small, reviewable steps that preserve public behavior.
• Verification: narrow tests, type checks, builds, or smoke checks that prove the split.

Lead each plan with one explicit verdict: ### ✂ Split justified, ### 🧱 Keep intact, or ### ⚠️ Generated — change the generator. Use a compact evidence table for repeated criteria. When proposing a split, show the source and target modules as a tree only when it clarifies ownership.

If a selected file is generated or vendored because --include-generated was used, plan against the generator, schema, or upstream source instead of hand-splitting generated output.

Guard Rails

• Treat the table as the source of truth for size ranking only. Rank refactor priority separately.
• Keep the plan cohesive. Do not split solely to reduce line count.
• Prefer existing project module boundaries and naming conventions.
• Report when the helper used its portable LOC estimate instead of tokei.

Completion

Complete with the exhaustive threshold report plus evidence-ranked plans only for candidates whose cohesion or change risk justifies a split. Lead with ### 🔎 Large-file scan — <candidate count>, state when a large file should remain intact and why, and surface the portable-LOC fallback as ⚠️ Approximate counts when used. Keep the helper's exhaustive table, paths, LOC values, and commands exact and undecorated.