Claude Skills by Z-M-Huang

Start the multi-AI pipeline. Plan -> Review -> Implement (loop until reviews approve). Codex final gate.

Build stage — per-unit implementation with fresh context and mechanical backpressure

PK Stage — multi-AI competitive debate with iterative consensus. Sends a topic to all configured AI participants simultaneously, synthesizes the best approach, then iterates until consensus or max rounds.

Code review stage — multi-AI semantic drift detection with AC tracing

Dev Buddy web configuration portal for managing presets and pipeline config

Decomposition stage — break features into small, independently testable units of work

Discovery stage — multi-AI codebase and running app exploration

Run a single task using a specific AI provider and model. Supports subscription, API, and CLI presets.

Pre-build validation — red-test check + Contract Manifest wiring check

Drive a Ralph run end-to-end through the dev-buddy MCP server.

Requirements + UAT design stage — acceptance criteria and Playwright test scenario authoring

UAT stage — execute Playwright tests and mechanical backpressure against the running app

Create a custom MCP tool for the documentation manifest. Guides the user through defining the tool's purpose, generating an action script with embedded data, computing the hash, and adding it to the manifest.

Generate AI-written documentation for a specific directory or all undocumented directories. Creates structured READMEs optimized for both human and AI consumption, and updates the manifest.

Initialize documentation manifest for git-doc-mcp. Scans the project for documentation files (*.md, READMEs, ADRs, etc.), creates .mcp/manifest.yml with resource entries and search tools, and configures .mcp.json for the git-doc-mcp MCP server. Run this once when setting up project documentation.

Scan project and report documentation coverage. Finds undocumented directories, stale manifest entries, and prioritized recommendations.

Sync the git-doc-mcp manifest with current project state. Adds new docs not yet indexed, removes stale entries, refreshes descriptions and search index, and regenerates tool actions.

Run a comprehensive audit against all applicable VCP standards. Supports full audit, compliance-specific audit, and quick release readiness check.

View and modify VCP configuration. Add or remove ignore entries, toggle scopes, manage compliance frameworks, change severity threshold, and manage exclude patterns. Supports both project config (.vcp/config.json) and global config (~/.vcp/config.json).

Inject VCP security and architecture standards into context. Run this at session start or after context compaction so the AI internalizes rules while writing code.

Identify untested code by mapping source files to test files. Finds functions without tests and tests missing edge case coverage.

Verify project dependencies against VCP dependency management standards. Checks lockfile hygiene, version ranges, package existence, and suspicious packages.

Initialize VCP configuration for this project. Creates global config (~/.vcp/config.json) if it doesn't exist, then creates project config (.vcp/config.json). Detects frameworks, scopes, and discovers the plugin path. Run this once when setting up VCP for a new project.

Review staged/changed files against all applicable VCP standards before committing. Produces a PASS/BLOCK verdict. Run this before every commit.

Review test quality against VCP testing standards. Checks for tautological tests, over-mocking, missing edge cases, and other anti-patterns.

Analyze a proposed bug fix against VCP root cause analysis standards. Determines whether a fix addresses the root cause or just patches a symptom.

Generate a test plan for specific code based on VCP testing standards. Produces unit tests, integration tests, edge cases, and mock guidance.