The Unofficial Home Assistant AI Toolkit

Design and run focused acceptance scenarios to check MCP tools through real agent CLIs such as Claude or Gemini. Review task completion, tool failures, and detailed run output, and compare branches when investigating a regression.

• Validate MCP tool changes in a pull request
• Check whether a tool regression occurs from an agent’s perspective
• Run an end-to-end scenario with Claude or Gemini CLI

Finds authors of merged pull requests who are missing from a README contributors list, drafts concise descriptions of their work, and updates the list only after the user approves the proposed additions.

• Find recent merged PR authors missing from the contributor list
• Draft factual one-line summaries of contributors’ changes
• Update a README after reviewing proposed additions

Manage your own GitHub pull request by checking CI results, review comments, and unresolved threads. Triage feedback, make code fixes, reply to comments, resolve threads, and recheck until outstanding issues are cleared.

• Check CI and review feedback on your own pull request
• Triage bot and human comments and fix relevant issues
• Resolve review threads after replying with an update

Review an external GitHub pull request for security concerns, test coverage, change size, and alignment with its linked issue or stated purpose. It summarizes findings and drafts a constructive review comment for you to approve before posting.

• Review an external contribution before deciding whether it is ready to merge
• Check whether a pull request has tests for its changed code
• Assess whether a contributor’s changes match the linked issue

Compare MCP tool behavior in a target version with a released baseline. It selects relevant prebuilt scenarios, designs custom regression stories, runs them with agent CLIs, verifies outcomes, and reports behavior, tool usage, and cost trends.

• Compare tool behavior between a release and a code change
• Choose test stories that exercise modified MCP tools
• Create regression scenarios for changed code paths

Analyze a GitHub issue by exploring related code, checking relevant external details, and assessing implementation complexity and architectural choices. The skill drafts a structured issue comment and recommends labels for maintainers to review.

• Assessing the complexity of a GitHub issue
• Planning implementation approaches
• Preparing an issue analysis and labels

Take a GitHub issue through implementation in a worktree, tests, and a draft pull request. The skill then checks CI results and review threads, applies fixes, and repeats the review cycle until the pull request is ready for maintainer review.

• Implementing a GitHub issue end to end
• Creating a draft pull request from an issue
• Resolving CI failures and review comments

Creates a Git worktree under `worktree/` for a named branch, after updating `master`. It initializes submodules and confirms the new location and branch, giving you an isolated checkout for separate work.

• Create an isolated checkout for a new feature branch
• Start parallel work from the latest master
• Initialize submodules in a new worktree