tednaleid

Audit and set up CI, release, Justfile, and Homebrew infrastructure for any repo. Detects language (Rust, Zig, Swift, TypeScript/Bun) and project type (CLI, macOS app) automatically. Use this whenever the user mentions CI setup, release automation, homebrew tap, justfile recipes, bump/retag commands, pre-commit hooks, or wants to normalize build infrastructure across projects. Also use when the user asks about GitHub Actions workflows for building, testing, or releasing their project.

Deep multi-lens review of an MR, PR, or any local branch. Dispatches 4 parallel subagents (architecture, security, test coverage, naming/API) inside an isolated worktree, aggregates findings with severity tags, and writes a dark-themed HTML disposition tracker (with checkboxes and localStorage) plus copy-paste comment drafts to .llm/reviews/. Nothing auto-posts; the human reviews and posts by hand. Use when asked to deeply review an MR/PR, do a thorough code review of a branch, get a second opinion on changes, or uses /review-branch. Works against GitLab (glab) and GitHub (gh); auto-detects from the git remote. Also handles local-only branch reviews (no MR/PR yet) by diffing against the default branch.

Create a git worktree for reviewing an MR/PR or working on a branch in isolation. Use when the user wants to review a merge request or pull request, check out a branch in a separate worktree, or uses /worktree. Accepts an MR/PR number, an MR/PR URL, or a branch name. Creates the worktree under .claude/worktrees/, copies gitignored environment files from the main worktree, bootstraps the project, and switches the session into it.

Serialize a long Claude Code session's working state into a relay markdown doc, iteratively refine it by consulting a fresh-context reviewer subagent, and emit a copy-paste resume prompt for a brand-new Claude instance. Use whenever the user mentions context getting full, wanting to hand off a session, before running /clear, before compaction, passing the baton to a fresh Claude, checkpointing current work, saving session state, or asking for a resume prompt. Also use proactively when the user notices the conversation is getting long and valuable context would be lost if the session ends.

Survey a codebase and write ONBOARDING.md (under 100 lines) covering stack, build/test/lint commands, architecture, entry points, and CI/CD. Use when the user asks to be onboarded to a repo, oriented to a new project, or wants their onboarding docs refreshed. Lives beside CLAUDE.md and loads on demand, so orientation doesn't bloat every turn.

Use this skill when working with veer - the PreToolUse hook that rewrites or blocks Bash tool calls in this repo. Triggers when the user edits .veer/config.toml, asks to "block" or "stop the agent from running" a command, wants to redirect calls like pytest/npm/cargo to a Justfile target, or mentions veer, rules, or PreToolUse hooks. Also use proactively: when the repo has a Justfile/package.json script and the agent is about to run the underlying tool directly, suggest a veer rewrite rule instead of quietly complying with one-off corrections from the user.

Interactive testing of the running montty app via its debug HTTP server. Use this skill whenever you need to verify app behavior yourself -- after implementing a feature, fixing a bug, or when the user asks you to "try it", "test it", "check if it works", "run the app and see", or "exercise it". Also use when debugging session persistence, tab state, split layouts, or any visual/behavioral issue that can't be verified by unit tests alone.

Inspect the running Limn desktop app via its debug HTTP server. Use this skill whenever you need to check what the app is rendering, verify DOM state, look for JS errors, capture screenshots, check node counts, or investigate any runtime behavior of the desktop app. Also use when debugging issues that require seeing what the WKWebView is actually displaying. Trigger on phrases like "check the app", "what does it look like", "is it rendering", "any JS errors", "screenshot the app", "how many nodes", or any ad-hoc investigation of the running desktop app.

Use this skill when working with veer - the PreToolUse hook that rewrites or blocks Bash tool calls in this repo. Triggers when the user edits .veer/config.toml, asks to "block" or "stop the agent from running" a command, wants to redirect calls like pytest/npm/cargo to a Justfile target, or mentions veer, rules, or PreToolUse hooks. Also use proactively: when the repo has a Justfile/package.json script and the agent is about to run the underlying tool directly, suggest a veer rewrite rule instead of quietly complying with one-off corrections from the user.

Use this skill when working with veer - the PreToolUse hook that rewrites or blocks tool calls in this repo. Triggers when the user edits .veer/config.toml, asks to "block" or "stop the agent from running" a command, wants to redirect calls like pytest/npm/cargo to a Justfile target, wants to enforce content rules on plans (ExitPlanMode), or mentions veer, rules, or PreToolUse hooks. Also use proactively: when the repo has a Justfile/package.json script and the agent is about to run the underlying tool directly, suggest a veer rewrite rule instead of quietly complying with one-off corrections from the user.

Create a git commit with an issue/JIRA ticket prefix and succinct message. Use when the user asks to commit changes, create a commit, or uses /commit. Determines the Issue/JIRA ticket from the branch name, stages specific files, and commits with a properly formatted message.