Import skills from external repositories into this Agent Skills repository. Use when pulling in a public skill, reviewing an external skill set, or adapting upstream content into an existing skill.
Keep the user in the loop through the Plannotator Inbox during long autonomous work, with a status update for every step, directions as gated Plannotator reviews and questions as Inbox questions, so they can follow and steer without the terminal. Use ONLY as an explicitly invoked OpenCode `/goal` objective alongside a loop skill and a review mode, such as `/goal @issue-loop @inbox-reporting @review-steering`.
Work through a GitHub issue and every issue related to it until they are done, committing each change and pushing at each working point, with the issue tracker, any project board and the todo list kept current and the user's steering folded in as it arrives. Use ONLY as an explicitly invoked OpenCode `/goal` objective naming the issue, such as `/goal @issue-loop` followed by its number, usually alongside `inbox-reporting` and one review mode (`review-stages` or `review-steering`).
Lit rendering and picker callback-shape guidance. Use when editing or reviewing Lit component render logic, derived render data, memoization, or picker callbacks.
Bring docstrings, comments, docs sites and external docs repositories up to date with recent code changes. Use when asked to update, check or catch up documentation.
Prepare a brand-new npm and JSR package for its first release from GitHub Actions. The bundled script publishes an npm 0.0.0 placeholder, adds the npm trusted publisher, and creates and links the JSR package. Use before a new package's first release, or when its release fails because the npm package doesn't exist, has no trusted publisher, or the JSR package isn't linked.
Develop and migrate OpenCode V2 plugins, clients, SDK hosts, and HTTP API integrations. Use for the OpenCode plugin API, `@opencode/client`, `@opencode/sdk`, server API, Effect entrypoints, or V1-to-V2 API migration.
Start, stop, check and tail local dev servers through the project's workflow, pitchfork or a framework background mode, and set up `pitchfork.toml` with stable .localhost addresses. Use for dev or background servers, pitchfork tasks, port clashes or unresolved proxy URLs.
Produce implementation-ready plans from the current conversation and repository context. Use when entering native plan mode, invoking /plan, or when a task needs concrete implementation sequencing before edits begin; do not use for round-based grilling.
Create and publish a major, minor, or patch release of @timmo001/oxlint-rules. Use when asked to create, cut, prepare, or publish an oxlint-rules release, including its version bump and npm and JSR publication.
Snapshot and diff changes to live systems outside Git, such as Home Assistant, servers over SSH, containers, databases or cloud APIs. Use before changing remote config or state, and when asked what was changed there.
Safe inlining and removal of single-use functions during code review and refactoring. Use when a local, non-exported helper has exactly one real call site and inlining preserves behaviour and readability.
Investigate a topic against primary sources and return cited findings. Use when the user asks why, wants evidence or validation, needs docs, API, spec or library behaviour verified, or compares competing views.
Run a retrospective on one or more agent sessions and propose improvements to the agent's environment, including skills, steering files, automated checks, and tooling. Use ONLY when the user invokes /retro or explicitly asks for a retrospective or a review of past sessions.
Review mode for long autonomous work where each stage gets a gated Plannotator code review before it is pushed, with feedback fixed and re-reviewed while independent work carries on. Use ONLY as an explicitly invoked OpenCode `/goal` objective alongside a loop skill, such as `/goal @issue-loop @inbox-reporting @review-stages`; use `review-steering` instead when the user only wants to steer.
Review mode for long autonomous work whose design is already settled, with no review requests, sign-offs or design questions. Changes are pushed as they reach a working point, decisions are made from the given design and reported, and the user steers only when they choose to. Use ONLY as an explicitly invoked OpenCode `/goal` objective alongside a loop skill, such as `/goal @issue-loop @inbox-reporting @review-steering`; use `review-stages` instead when each stage needs sign-off.
Safe process killing and signal handling for agent/subprocess contexts. Use when running pkill, killall, kill, or any process termination command from a shell subprocess, automated script, or coding agent.
Split independent work into visible Herdr sessions and choose models and effort variants for each assignment. Use when tasks benefit from parallel workers or different model capabilities, when the user requests coordination or model selection, or when a coordinator agent is selected. Ask before launching workers.
Use, configure, maintain, or create reusable GitHub Actions workflows for personal and organisation repositories. Use when a task mentions shared workflows, reusable workflows, `workflow_call`, cross-repository workflow `uses:`, or the personal workflows repository; do not use for repository-specific or proof-of-concept CI unless evaluating whether it should be shared.
A folder with a SKILL.md file: a name, a description of when to use it, and instructions. Claude loads a skill only when the task matches its description.
How do I use one I find here?
Copy the folder into your project's .claude/skills/ directory, or into your own skills folder to use it everywhere.
What do the warnings mean?
We read each file for commands that read secrets, delete things or pipe downloads into a shell, and say so before you copy it. No warning is not a promise that a file is safe.
Which skills worked for people?
Open a skill to see its discussion. Reports from people and their agents are coming.