| name | review-duplicates |
| description | Duplicate code detection using similarity-rs. Identifies refactoring candidates based on semantic similarity. |
| user-invocable | true |
| context | fork |
| agent | code-refactor-agent |
When subagents are enabled in Codex, prefer delegating this workflow to the code-refactor-agent custom agent.
Detect duplicate code in the ccswarm codebase using similarity-rs.
similarity-rs crates/ --threshold 0.80
For Rust code, also consult cargo coupling so duplicate cleanup is prioritized
by architectural impact, not only by textual similarity.
cargo coupling . --summary --exclude-tests
cargo coupling . --hotspots=12 --exclude-tests
If a duplicate cluster is located in a coupling hotspot, high-risk boundary, or
cycle-related module, raise its refactoring priority accordingly.
Classify results by priority:
| Priority | Condition | Action |
|---|
| High | 95%+ similarity, 10+ lines | Extract to common function |
| Medium | 90-95% similarity, 5+ lines | Consider trait or macro |
| Low | 85-90% similarity | Evaluate structural similarity |
| Skip | Test code, config structs | Allow intentional duplication |
Output a JSON report with high_priority, medium_priority, low_priority, and ignored sections.