Skip to main content

navvy

Diagnose a codebase through its git history before reading any code. Surfaces churn hotspots, bus factor risks, bug clusters, velocity trends, and firefighting patterns. Only invoked directly via /navvy — does not auto-trigger.

Ir a la instalación

Datos de origen

Repositorio
andrewgleave/skills
Última actividad en el origen
8 de abril de 2026 a las 14:31
Idioma detectado de SKILL.md
inglés
Estrellas
33
Forks
1

Opciones de instalación

De forma predeterminada está seleccionado el prompt que primero revisa el origen. Puedes cambiar a un comando directo o descargar una copia local.

Revisa los archivos de origen

Lee SKILL.md y los archivos complementarios que muestra SkillsMP antes de decidir si quieres instalarlo.

Explorador de archivos
2 archivos

Mostrando SKILL.md

SKILL.md
Instrucciones de origen · Vista previa de solo lectura
name
navvy
description
Diagnose a codebase through its git history before reading any code. Surfaces churn hotspots, bus factor risks, bug clusters, velocity trends, and firefighting patterns. Only invoked directly via /navvy — does not auto-trigger.
# Navvy Five git commands, run in under a minute, that tell you more about a codebase than an hour of reading code. Commit history is diagnostic data — it reveals team dynamics, risk concentration, and maintenance patterns that the code itself cannot. Run all five commands, then synthesize the results into a diagnostic report. ## The Commands Run these in parallel where possible. Adjust `--since` to match the project's age — use `1 year ago` for active projects, extend for slower-moving ones. ### 1. Churn hotspots ```bash git log --format=format: --name-only --since="1 year ago" | sort | uniq -c | sort -nr | head -20 ``` The 20 most-modified files in the past year. High churn on a file that nobody wants to own is the clearest signal of codebase drag — these files have unpredictable blast radius and inflate estimates. ### 2. Contributor map ```bash git shortlog -sn --no-merges ``` Ranks contributors by commit count. Look for bus factor risk: if one person accounts for 60%+ of commits, knowledge is dangerously concentrated. Also check whether core contributors are still active. ### 3. Bug clusters ```bash git log -i -E --grep="fix|bug|broken" --name-only --format='' | sort | uniq -c | sort -nr | head -20 ``` Files most frequently touched in bug-related commits. Files appearing in both this list and the churn hotspots are highest-risk code — they keep breaking and keep getting patched. ### 4. Velocity trend ```bash git log --format='%ad' --date=format:'%Y-%m' | sort | uniq -c ``` Commit counts by month across the full history. A steady rhythm is healthy. Sharp declines suggest departures or loss of momentum. Sporadic spikes suggest batched releases rather than continuous delivery. ### 5. Firefighting signals ```bash git log --oneline --since="1 year ago" | grep -iE 'revert|hotfix|emergency|rollback' ``` Reverts and hotfixes. Frequent occurrences signal the team doesn't trust its deploy process — a symptom of deeper issues with testing or deployment reliability. ## Synthesizing the Report After running all five commands, write a short diagnostic report. Structure it around what the data reveals, not around the commands themselves. ### What to look for - **Overlap between churn and bugs.** Files appearing in both lists are the highest-leverage targets for refactoring or better test coverage. - **Bus factor.** A single dominant contributor who's gone inactive is a risk. A well-distributed contributor map is healthy. - **Velocity shifts.** Correlate drops or spikes with what you know about the project — team changes, rewrites, launches. - **Firefighting frequency.** A few reverts a year is normal. Monthly hotfixes suggest systemic issues. ### Report format ``` # Codebase Diagnostic: [repo name] ## Key Findings [2-4 bullet points: the most important things to know before touching this code] ## Churn Hotspots [Top files, what they are, whether they overlap with bug clusters] ## Team & Ownership [Bus factor assessment, active vs inactive contributors] ## Health Signals [Velocity trend interpretation, firefighting frequency] ## Recommendations [Where to focus attention, what to investigate further] ``` Keep it concise. The point is orientation, not exhaustive analysis. A reader should finish the report knowing where the bodies are buried and where to tread carefully.
Ver en GitHub