rebuild

Rebuild Project Knowledge

Scan the entire codebase and generate a complete project knowledge base from scratch.

Announce at start: "I'm rebuilding the project knowledge base from the codebase."

When to Use

• First time setting up the knowledge base for a project
• Knowledge base has drifted significantly from reality
• User explicitly requests a full rebuild

Process

1. Scan the codebase:

   • Read project structure: ls, key directories, entry points
   • Read configuration files: package.json, Cargo.toml, pyproject.toml, Makefile, docker-compose.yml, etc.
   • Read existing documentation: README.md, CLAUDE.md, docs/ directory
   • Read existing specs and plans: docs/superpowers/specs/, docs/superpowers/plans/
   • Check git log for recent development history: git log --oneline -20
   • Sample key source files to understand architecture patterns

2. Generate knowledge files:

   Create docs/project-knowledge/ directory if it doesn't exist.

   For each of the 5 knowledge files, use the plugin template as the structural basis and fill in concrete content from the codebase analysis:

   • architecture.md — System overview, module structure (from directory layout and imports), data flow (from entry points and key modules)
   • tech-stack.md — Languages and frameworks (from config files), key dependencies (from package manifests), build tools (from scripts/Makefile)
   • features.md — Implemented features (from specs, README, and code), in-progress features (from plans with unchecked items)
   • conventions.md — Coding standards (from linter configs, existing patterns), architecture rules (from CLAUDE.md or observed patterns), testing conventions (from test directory structure and existing tests)
   • decisions.md — Extract significant decisions from git history, specs, and code comments. Create ADR entries for non-obvious architectural choices.

3. Set frontmatter: For every generated file:

   • last_updated: today's date (YYYY-MM-DD)
   • updated_by: superpowers-memory:rebuild
   • triggered_by_plan: "none" (literal string — not YAML null)

4. Generate MEMORY.md index:

   After writing the 5 knowledge files, generate docs/project-knowledge/MEMORY.md:

   • For each of the 5 files, extract 2-3 concrete key points from the content you just wrote (e.g., specific pattern names, version numbers, rule names — not generic descriptions)
   • Write the file following the format in templates/MEMORY.md, setting updated_by: superpowers-memory:rebuild and triggered_by_plan: "none"

   Size constraint: Keep MEMORY.md under 30 lines total.

5. Commit:

git add docs/project-knowledge/
git commit -m "docs: rebuild project knowledge base from codebase"

6. Report:
   • Summarize what was generated
   • Note any areas where information was sparse or uncertain
   • Suggest running superpowers-memory:update after the next iteration to keep knowledge fresh

Quality Guidelines

• Be factual: only include what you can verify from the codebase. Do not speculate.
• Be concise: each file should be scannable in under 2 minutes
• Be structured: follow the template format strictly so incremental updates work cleanly
• Link to sources: reference file paths, spec files, and plan files where relevant