Smithery is now a part of Arcade.dev! Read more in our announcement here
Analyze Git history to uncover “logical coupling” (files that always change together) and “hotspots” (frequently modified, complex modules). Based on Adam Tornhill’s Your Code as a Crime Scene.
“History doesn’t repeat itself, but it rhymes. Static analysis tells you structure; Git forensics tells you the painful truth.” — Adam Tornhill
This skill is based on Adam Tornhill’s Your Code as a Crime Scene methodology. Core idea: a system’s evolution history reveals design problems better than the code alone.
[!IMPORTANT] Before performing any analysis, you must invoke the
mcp_sequential-thinking_sequentialthinkingtool and reason for 3–10 steps, or more if needed.Example thinking prompts:
- “How deep is this project’s Git history? Do I need
git fetch --unshallow?”- “Which time window should I focus on? (Last 3 months? 1 year?)”
- “Are there obvious noise files (e.g.
package-lock.json) that must be excluded?”
Coupling analysis
python scripts/git_forensics.py --repo . --threshold 0.3
Hotspot detection
python scripts/git_hotspots.py --repo . --days 180
Master’s saying: “The value of code is not what it is, but how it became that way.”
Run:
git log --oneline -n 100
to quickly gauge recent project activity.
Key inference: If the last ~50 commits touch only one or two directories, that’s the epicenter—the area most likely hiding risk.
(Temporal Coupling / Change Coupling)
Core question: “Are there two files that have no import/use relationship, yet appear together in 70% of commits?”
Master warnings (from Adam Tornhill):
⚠️ Logical coupling with physical separation → strong signal of architectural decay
⚠️ Cross–build-root coupling → if service/ipc.rs and gui/api.ts change together but belong to different build roots, this is a breeding ground for version skew
Prescription:
(CodeScene Methodology)
Hotspot = High Change Frequency (Churn) × High Complexity| Low Complexity | High Complexity | |
|---|---|---|
| High Churn | Config / generated code — often ignorable | 🔴 Top refactor priority (bug breeding ground, highest ROI) |
| Low Churn | Stable modules — leave them alone | 🟡 Legacy minefield — tread carefully |
git fetch --unshallow. No history = no data = blind analysis.package-lock.json, Cargo.lock, *.min.js, dist/, and other generated artifacts—they pollute results.git mv can distort coupling signals. If results look strange, manually verify rename history.Your report must include:
