context-engine

Audit the project against the plugin's prescribed standards and refactor the project to conform. Run after `/init` to identify gaps, then apply fixes per dimension.

Mark a feature as CANCELLED, capture learnings from partial work, and optionally clean up branches and checkpoints.

Create, list, rollback, and clean hybrid checkpoints (git tag + .context/ snapshot).

Create a new skill with proper structure, scoped content, and a quality review pass. Adds the skill to the framework's skill table and updates the skill profile.

Something is broken. Uses parallel hypothesis testing via Agent Team for complex bugs, or single `debugger` subagent for simple ones.

Display framework health dashboard and run deep analysis. Reads from `.context/metrics/HEALTH.md`, git history, and `.context/` state.

Execute the PRP using an Agent Team for parallel step execution.

Bootstrap the context engineering system. Run this first on any new or existing project. Analyzes the codebase and populates `.context/`.

Search, browse, and maintain the deep knowledge layer.

Capture a learning from the current session. Routes to the correct file based on type.

Streamlined for bug fixes, single-file changes, minor improvements. Stays in one context.

No code in this phase. Create a PRP that guides implementation.

Restructure code without adding features. Uses Agent Team for multi-module refactors or single subagent for focused changes.

No code in this phase. Understand what exists before planning.

Reload relevant context efficiently after `/clear` or new chat.

Run a focused security review on specified files or recent changes. Use ad-hoc when you want a security assessment outside the `/validate` workflow.

Identify and apply code simplifications: dead code, duplication, over-abstraction, and consolidation opportunities. Use ad-hoc when you want a focused cleanup pass outside the `/validate` workflow.

Synthesize the current state of the project from `.context/`. Useful for onboarding, catching up after a break, or getting the big picture.

Update `.context/architecture/` after significant codebase changes. Called by `/validate` when structural changes are detected, or manually after major refactors.

Refresh one or more skills by diffing their current content against what the project actually does today. Pulls from `.context/knowledge/`, live source files, and dependency manifests to keep skills accurate and actionable.

Mandatory regardless of testing strategy. Uses an Agent Team for parallel review when changes are substantial.

Systematically diagnoses and resolves bugs. Uses diagnostic reasoning, log analysis, hypothesis testing, and bisection. Use when something is broken and needs investigation.

Executes PRP steps following the project's testing strategy. Works sequentially, validates after each step.

Executes MCP tool operations in isolated context and returns compressed summaries. Use for browser automation, database queries, documentation lookups, and other MCP-heavy tasks.

Creates structured PRPs (Product Requirements Prompts) from research notes. Produces actionable plans with file paths, test coverage needs, and validation criteria.

Explores codebase structure, finds relevant files, and returns distilled summaries. Use before planning to understand existing code without consuming main context.

Reviews code changes for quality, pattern compliance, security, and edge cases. Use after implementation, before commit.

API conventions for this project. Auto-loaded when working with routes, controllers, and endpoint files to ensure consistent API patterns.

Authentication and security patterns - JWT, OAuth, sessions, RBAC, input validation. Auto-loaded when working with auth, middleware, or security files.

Chrome DevTools debugging and browser automation via MCP. Auto-loaded when working with browser debugging, performance analysis, or web inspection tasks.

Context Engine framework reference. Auto-loaded when working with .context/, agents, commands, or hooks. Supplements CLAUDE.md with role and skill details.

Library documentation lookup via Context7 MCP. Use when you need current docs for any library, framework, or API. Prevents hallucinated APIs by fetching real documentation.

Database migration patterns and ORM conventions. Auto-loaded when working with schema changes, migrations, models, or SQL files.

Deployment and CI/CD patterns - Docker, GitHub Actions, pipelines, infrastructure. Auto-loaded when working with deployment configs, workflows, or container files.

Git workflow conventions - branching, commits, PRs, rebasing. Auto-loaded when working with git operations, merge conflicts, or PR-related files.

Google Workspace operations via MCP - Docs, Sheets, Drive. Use when creating, reading, or editing Google Docs, Sheets, or Drive files from Claude Code.

Deep knowledge layer - project-specific library quirks, stack recipes, dependency notes. Auto-loaded when working with knowledge files or when implement/debug discovers something worth preserving.

MCP server integration catalog and configuration reference. Auto-loaded when working with MCP configs. Lists all available and recommended MCP servers with install commands.

Node.js backend patterns - Express, Fastapi, NestJS, middleware, async. Auto-loaded when working with server.ts/js, app.ts/js, routes/, middleware/, or Node.js backend directories.

PostgreSQL database operations via MCP - schema inspection, live queries, EXPLAIN analysis. Use when you need to inspect the actual database state, verify migrations, or analyze query performance.

PostgreSQL patterns - queries, indexing, performance, maintenance. Auto-loaded when working with SQL files, database configs, or Postgres-specific code.

Context budget and token efficiency rules. Auto-loaded to minimize prompt overhead across all interactions.

Browser automation via Puppeteer MCP - navigation, screenshots, form filling, testing. Auto-loaded when working with browser tests, E2E tests, or scraping-related files.

Python backend patterns - FastAPI, Django, Flask, project structure, async. Auto-loaded when working with Python backend files.

React and frontend patterns - components, hooks, state management, performance. Auto-loaded when working with React/JSX/TSX files or frontend directories.

Redis patterns - caching, sessions, queues, pub/sub, rate limiting. Auto-loaded when working with cache, Redis, or queue-related files.

Ruby and Rails patterns - models, controllers, services, testing. Auto-loaded when working with Ruby files, Gemfiles, or Rails directories.

Sequential thinking for structured problem decomposition. Use for complex planning, debugging, and architectural decisions where step-by-step reasoning with revision and branching helps.

Testing conventions for this project. Auto-loaded when working with test files to ensure consistent test patterns.

TypeScript patterns and type safety. Auto-loaded when working with .ts files, tsconfig.json, or type definition files (.d.ts).

12 hooks across 6 events