Skip to content

im-shashanks/claude-plugins / shaktra/skills/shaktra-analyze

shaktra-analyze

Verified

Codebase Analyzer workflow for brownfield codebases — deep structural analysis producing structured YAML artifacts for downstream agents. Use this skill whenever the user wants to analyze, assess, or understand an existing codebase — including requests like "analyze my codebase", "what's the health of this project", "check for tech debt", "audit dependencies", "understand this codebase before we start", or any brownfield due diligence task.

FileKindSize
SKILL.mdskill md13.4 KB
analysis-dimensions-core.mdmarkdown14.2 KB
analysis-dimensions-git.mdmarkdown5.5 KB
analysis-dimensions-health.mdmarkdown12.7 KB
analysis-output-schemas.mdmarkdown9.2 KB
debt-strategy.mdmarkdown5.7 KB
deep-analysis-workflow.mdmarkdown13.2 KB
dependency-audit.mdmarkdown6.1 KB
standard-analysis-workflow.mdmarkdown7.2 KB

analysis-dimensions-git.md

View on GitHub
# Analysis Dimensions — Git Intelligence (D9)

Historical signals from git that are invisible to static analysis. Answers "How has this codebase EVOLVED?" — change patterns, bug density, knowledge distribution, and code age.

**Used by:** shaktra-cba-analyzer agent during `/shaktra:analyze` workflow.
**Input dependency:** Requires a git repository. Consumes `static.yml` (file inventory for scoping).
**Companion:** `analysis-dimensions-core.md` defines D1-D4, `analysis-dimensions-health.md` defines D5-D8.

---

## Extraction Philosophy

All data in this dimension comes from **tool-based extraction** — the same philosophy as `static.yml`. Run git commands via Bash and parse the output. No LLM interpretation of git data. The CBA analyzer's role is structuring the output, not inferring meaning.

**Shallow clone guard:** Before extraction, check commit count with `git rev-list --count HEAD`. If fewer than 10 commits, note the limitation in `summary:` and produce minimal output (hotspots and code_age only, with a caveat).

---

## D9: Git Intelligence → `git-intelligence.yml`

**Scope:** Historical change patterns, bug-fix density, co-change coupling, knowledge distribution, and code age across all tracked source files.
**Downstream impact:** The cross-cutting risk correlation (D8) uses change frequency and bug-fix density as key risk factors. The sw-quality agent uses hotspots and bus factor to identify files needing extra review attention. Co-change patterns reveal hidden coupling that static analysis misses — the architect uses these to validate or challenge module boundary decisions.

**What to extract:**

1. **Hotspots** — files with the most commits. Use `git log --format=format: --name-only | sort | uniq -c | sort -rn` scoped to source files from `static.yml`. Classify: top 10% = hot, middle 40% = warm, bottom 50% = cold. Also extract distinct author count per file via `git shortlog -sn -- {file}`.

2. **Bug-fix density** — commits matching fix/bug/patch/hotfix patterns in their message. Use `git log --oneline --all -- {file}` and grep for `fix|bug|patch|hotfix` (case-insensitive). Compute fix_ratio = fix_commits / total_commits. Also count recent fixes (last 90 days) via `git log --since="90 days ago"`.

3. **Co-change patterns** — files that frequently change together in the same commit. Use `git log --format=format: --name-only` to get file lists per commit, then compute pairwise co-occurrence. Report clusters with minimum 3 co-occurrences. Flag `hidden_coupling: true` when co-changing files don't import each other (cross-reference `static.yml` dependency graph).

4. **Knowledge distribution** — author concentration per file. Use `git shortlog -sn -- {file}` for commit counts per author. Compute bus_factor (authors with >20% of commits). Flag `knowledge_risk: high` when bus_factor = 1 and file is on a critical path.

5. **Code age** — last modification date per file. Use `git log -1 --format=%aI -- {file}`. Classify: fresh (<30 days), recent (30-180 days), aging (180-365 days), stale (>365 days). Age categories are analysis constants, not quality gates.

**Git commands reference:**
- Hotspot ranking: `git log --format=format: --name-only --diff-filter=ACDMR | sort | uniq -c | sort -rn`
- Per-file authors: `git shortlog -sn -- {file}`
- Fix-commit detection: `git log --oneline --all --grep='fix\|bug\|patch\|hotfix' -i -- {file}`
- Recent fix commits: `git log --oneline --since="90 days ago" --grep='fix\|bug\|patch\|hotfix' -i -- {file}`
- Co-change extraction: `git log --format=format: --name-only` (group by blank-line-delimited commits)
- Code age: `git log -1 --format=%aI -- {file}`
- Commit count check: `git rev-list --count HEAD`

**Evidence requirements:**
- All data from actual git command output — no LLM inference
- Hotspot thresholds based on actual commit count distribution (percentile-based)
- Co-change patterns require minimum 3 co-occurrences to report
- Bug-fix matching uses commit message pattern matching, not code analysis
- Knowledge distribution uses commit count attribution (approximation, not perfect)

**Output structure:**
```yaml
summary: |
  {Self-contained 400-token overview: hotspot count, highest-churn files,
   co-change clusters, bug-fix density distribution, knowledge risks}
details:
  hotspots:
    - file: string
      commits_last_90_days: integer
      commits_last_365_days: integer
      distinct_authors: integer
      churn_category: hot | warm | cold
  bug_fix_density:
    - file: string
      total_commits: integer
      fix_commits: integer
      fix_ratio: float
      recent_fixes: integer
  co_change_patterns:
    - cluster_name: string
      files: [string]
      co_change_frequency: integer
      linked_in_code: boolean
      hidden_coupling: boolean
  knowledge_distribution:
    - file: string
      primary_author: string
      author_count: integer
      bus_factor: integer
      knowledge_risk: high | medium | low
  code_age:
    - file: string
      last_modified: ISO-8601
      age_category: fresh | recent | aging | stale
```

**Self-check before writing:**
- All data comes from actual git command output — no numbers inferred or estimated
- Hotspot thresholds are percentile-based on the actual commit distribution — not arbitrary cutoffs
- Co-change patterns have minimum 3 co-occurrences — don't report noise from single shared commits
- If commit count is <10 (shallow clone), this is noted in summary and only minimal output is produced
- Knowledge distribution uses commit count attribution — acknowledge this is an approximation in the summary
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall