ai-harness-toolkit

Build a step-by-step path from a plain repository to a full agent-driven engineering workflow. Inspects the repo first to detect the full tech stack, package managers, CI systems, and monorepo layout before making any changes. Then initializes APM, sets up spec-driven development (OpenSpec by default, or respects existing systems like SpecKit and BMAD), adds deterministic checks matched to the detected stack, installs applicable AI skills, creates workflow documentation, configures AI IDEs (Copilot, Cursor, Windsurf, Claude Code, OpenCode), and verifies the entire setup.

Evaluate and score how AI-native a repository is. Checks for the existence and maturity of AI primitives -- spec-driven development, deterministic checks, AI skills, documentation, IDE configuration, and agentic legibility -- and produces a scored markdown report with prioritized recommended actions.

Stack-aware code review orchestrator that detects tech stacks from changed files, dispatches parallel review lanes to subagents with prescriptive instructions, synthesizes findings, and runs a fix loop for deterministic fixes. Supports TypeScript/JavaScript, Python, and Java.

Commit and push changes with exclusions and a generated commit message

Monitor a GitHub pull request after creation, wait for new review comments, resolve code review comments after fixes, and report CI success or failure. Use when the user wants the backend of the review flow handled after a PR is opened.

Create a pull request with automatically recommended reviewers based on git history. Use this skill when the user wants to create a PR with reviewers, open a pull request, or submit changes for review with the right people assigned.

Use for workflows that fetch GitHub pull request review comments with the gh CLI, identify Copilot review threads, and resolve them after code changes. Trigger when asked to review/resolve Copilot PR comments, fetch PR review threads, or document how to close review threads via gh api graphql.

Review and improve Python code using focused guidance on common anti-patterns, performance bottlenecks, test quality, and error handling. Use this skill when writing or reviewing Python code that needs judgment beyond linters and type checkers.

Analyze recent code changes, review findings, and touched docs to codify what changed in the repository. Use when work is complete, after a review/fix loop, or whenever the agent should update docs, commands, skills, and conventions based on recent changes while identifying repeatable problems that should become lint rules or other mechanical checks.