Stats
Actions
Tags
How this skill is triggered — by the user, by Claude, or both
Slash command
/skills:cleanup [module|all][module|all]The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Systematic cleanup of dead code, unused imports, and technical debt.
Systematic cleanup of dead code, unused imports, and technical debt.
RULE: All tests must pass after cleanup. No behavior changes.
$ARGUMENTS - Module to clean up, or all for entire codebaseLaunch these agents in parallel:
Agent A (architect): Run the full test suite and capture the output.
./gradlew test
Report pass/fail status and any existing failures. If tests do not pass, STOP. Do not proceed until all tests are green.
Agent B (architect): Run static analysis tools and capture current warning counts.
./gradlew detekt
./gradlew ktlintCheck
Report the total warning count and error count for each tool. These are the baseline numbers.
Wait for both agents to complete. If Agent A reports test failures, fix them before continuing. Record Agent B's baseline counts for comparison in the final summary.
All detection work is read-only analysis. Launch these agents in parallel:
Agent A - Unused Imports (Explore): Scan all .kt files in the target module(s) for unused imports. Use IDE-style analysis: for each import statement, search the file body for usage of the imported symbol. Produce a list of files with unused imports and the specific import lines.
Agent B - Dead Code (Explore): Search for dead code across the target module(s). Look for:
TODO/FIXME/HACK/XXX commentsFor each finding, note the file path, line number, symbol name, and why it appears dead.
Agent C - Deprecated Code (Explore): Search for all deprecation markers across the target module(s):
@Deprecated annotations@deprecated KDoc tags@DeprecatedFor each finding, note the file, the deprecated symbol, and whether a replacement is specified.
Agent D - Technical Debt (Explore): Search for technical debt indicators across the target module(s):
@Suppress annotations (note what warning is suppressed)For each finding, assess severity (High/Medium/Low) and estimated effort (Low/Medium/High).
Agent E - Code Style (Explore): Analyze code style consistency across the target module(s):
// ---, // ===, // ****) that should be removedAgent F - Stale Documentation (Explore): Scan for documentation issues across the target module(s):
// increment counter above counter++)@param/@return tags that don't match the current signatureWait for all six agents to complete. Collect all findings into a unified cleanup manifest organized by file.
./gradlew ktlintFormat
This handles the bulk of import cleanup automatically.
./gradlew test
Using the unified manifest from Phase 1, partition the affected files into groups. Use the consolidation pattern — launch each file group as a parallel worktree-isolated agent (isolation: "worktree"). Agents can freely edit overlapping files; the consolidator handles merges.
Each worktree agent: For its assigned file group, apply all queued fixes from the manifest:
@Deprecated annotation, migrate callers to the replacement. Where we own the deprecated symbol and it has zero external callers, remove it.// TODO(cleanup): [description] or create a GitHub issue.// ---, // ===), standardize error handling within each file.@param/@return tags.Each agent commits its changes before finishing.
After all worktree agents complete, launch the consolidator agent (subagent_type: "consolidator") to merge all branches and resolve any overlapping edits.
Launch a reviewer agent (subagent_type: "reviewer") to review the merged diff. Fix any critical/major issues found.
Launch a verifier agent (subagent_type: "verifier") in post-verification mode to confirm all tests pass, lint is clean, build succeeds, and no behavior changes occurred.
./gradlew ktlintFormat
Launch a verifier agent (subagent_type: "verifier") in post-verification mode. It runs tests, lint, detekt, and build. Compare warning counts against the Phase 0 baseline to quantify improvement.
Create cleanup report:
# Cleanup Report: [Module/All]
## Summary
- Files modified: X
- Lines removed: Y
- Warnings fixed: Z (before: B, after: A)
## Changes Made
### Unused Imports Removed
- X files had unused imports removed
### Dead Code Removed
- [file:symbol] - [why it was dead]
### Deprecated Code Migrated
- [file:symbol] - migrated to [replacement]
### Technical Debt Addressed
- [item] - [what was done]
### Code Style Fixed
- [description of patterns normalized]
### Documentation Cleaned
- [stale comments removed, KDoc added]
### Issues Created for High-Effort Items
- #NNN - [tech debt item]
## Remaining Items
- [items that need future attention]
Delegate to the skills:commit-all command to produce atomic, logically-grouped commits from the cleanup changes:
Skill(skill="skills:commit-all")
skills:commit-all will partition the diff into groups like unused imports, dead code removal, deprecation migration, style normalization, etc. and commit each group with a (chore): ... message.
npx claudepluginhub abnegate/claudes --plugin skillsCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.