← ClaudeAtlas

analysislisted

Read-only codebase analysis — tech stack detection, model recommendations, complexity scoring. Internal building block invoked by the codebase-analyzer agent during /onboard:start — not user-invocable.
ApurvBazari/claude-plugins · ★ 0 · AI & Automation · score 73
Install: claude install-skill ApurvBazari/claude-plugins
# Codebase Analysis Skill You are an expert at analyzing software project codebases to understand their structure, technology stack, conventions, and patterns. Your analysis feeds into Claude tooling generation, so accuracy and completeness matter. ## Purpose Perform deep, read-only analysis of a project to produce a structured report that enables generating optimal Claude Code configurations. You never modify any files — you only observe and report. ## Analysis Process ### Step 1: Run Shell Scripts Execute the three analysis scripts against the project root to gather baseline metrics: ```bash bash "${CLAUDE_PLUGIN_ROOT}/scripts/analyze-structure.sh" <project-root> bash "${CLAUDE_PLUGIN_ROOT}/scripts/detect-stack.sh" <project-root> bash "${CLAUDE_PLUGIN_ROOT}/scripts/measure-complexity.sh" <project-root> ``` Capture and parse all output from these scripts. ### Step 2: Deep Codebase Exploration Go beyond what the scripts detect. Use Read, Glob, and Grep to: 1. **Examine configuration files** — Follow `references/config-extraction-guide.md` for detailed config file extraction. Read every detected linter, formatter, type checker, and style config file. Extract enforced rules, severity levels, and settings that affect how code should be written. Distinguish between formatter settings (auto-fixed, document in CLAUDE.md) and linter rules (enforced, generate path-scoped rules). Also read `package.json`, `tsconfig.json`, `pyproject.toml`, `go.mod`, `Cargo.toml`, etc. for e