building-writing-kits
Build a writing kit, everything a person needs to write a doc themselves (sourced facts, verbatim quotes, the choices a draft made, reader objections, open questions, blanks only they can fill) without the finished prose. Use whenever the user wants material or "inspo" to write a brief, memo, RFC, PR/FAQ, proposal, review, or post in their own words, says they will write it themselves, asks for research or a packet to write from, wants an existing draft taken apart into what went into it, or wants their finished doc checked against the sources.
monitoring-ci
Watch a GitHub Actions run after a push and report pass/fail with the failing job logs. Use after any `jj git push` or `git push`, or when the user asks whether CI passed. Not for writing or debugging workflow YAML.
searching-history
Search past Claude Code sessions to recall prior solutions, commands, fixes, and decisions. Use when the user references earlier work ("how did I fix", "last time we", "pick up where we left off", "a few months ago", "what was that … I set up", "didn't I once").
specifying-and-planning
Plan project implementation -- extract requirements from a spec, break them into tasks, create GitHub issues and project boards, or generate an agent team configuration from the task list.
syncing-claude-config
Sync this config against the shipped Claude Code -- new settings, hooks, and env vars since the pinned baseline; schema drift on silently-validating surfaces (themes, settings); and instruction drift where CLAUDE.md, rules, or skills duplicate or contradict the product's own system prompt. Not for judging whether config is well-shaped (that is maintaining-claude-code).
using-jj
Jujutsu (jj) reference for anything past describe/new/commit/push -- the non-interactive split form, which revision a bookmark must target before pushing, absorb, evolog, oplog recovery, immutable heads, revsets. Load whenever a git habit could produce a wrong or editor-blocking jj command.
using-linear
Query, create, and manage Linear issues from the command line. Use when checking assigned work, viewing issue details, creating new issues, or updating issue status.
validating-project
Auto-detect and run formatters, linters, type checkers, and tests for the current project. Use when validating a project, running all checks, checking code quality before committing, or verifying the build passes.
analyzing-wav
Analyze WAV audio files for debugging TTS and audio pipelines. Use when inspecting sample rate, bit depth, channels, PCM values, waveform, or spectrogram of a WAV; validating WAV headers and format; comparing duration or amplitude between audio files; or diagnosing clipping, distortion, silence, and generated speech issues.
building-swiftui
Build, review, debug, and modernize SwiftUI apps for macOS with modern patterns. Use when building SwiftUI UIs, reviewing code quality, debugging view issues, checking anti-patterns, migrating from AppKit, or designing app architecture.
configuring-tmux
Build, review, and troubleshoot tmux configurations with modern best practices. Use when creating tmux.conf from scratch, adding plugins, configuring themes, integrating with Neovim, setting up session management, or debugging tmux issues.
cuda-kernels
Develop, test, and optimize custom CUDA kernels in the candle framework for qwen3-tts-rs. Use when writing CUDA kernels, debugging GPU code, integrating kernels with candle, optimizing throughput, profiling with nsys/ncu, analyzing roofline, or investigating register pressure and occupancy.
displaying-images
Display images in the terminal using the Kitty graphics protocol via kitty-img. Use when the user asks to show, display, view, preview, or render an image file (PNG, JPG, GIF, SVG, etc.) in the terminal, or when you want to visually present a generated image.
downloading-hf-models
Download models and weights from HuggingFace. Use when downloading models, fetching safetensors or GGUFs, running "hf download", cloning model repos, or pulling weights from HuggingFace.
implementing-issues
Implement a GitHub or Linear issue end-to-end with TDD, branch management, and PR creation. Use when implementing an issue, working on a ticket, starting a feature from an issue, or building from issue requirements.
managing-starship-fork
Sync, build, and install the custom starship fork with jj support. Use when updating starship, rebuilding the binary, syncing with upstream, or modifying the jj_status module.
profiling-qwen3-tts
Profile and benchmark qwen3-tts-rs inference using e2e_bench with chrome tracing, flamegraph, or Nsight Systems inside a CUDA Docker container. Use when profiling, benchmarking, checking performance, finding bottlenecks, or running flamegraphs.
reading-books
Read and search digital books (PDF, EPUB, MOBI, TXT). Use when answering questions about a book, finding quotes or passages, navigating to specific pages or chapters, or extracting information from documents.
testing-whisper
Transcribe WAV audio files using OpenAI Whisper for intelligibility testing. Use when transcribing audio, testing speech output quality, running whisper, or checking if generated audio is intelligible.
uploading-hf-models
Upload model weights to HuggingFace at high speed using best practices. Use when pushing models, uploading safetensors, uploading GGUFs, or creating model repos on HuggingFace.
using-1password
Create, retrieve, and share 1Password items using the op CLI. Use when creating secure notes, storing credentials, generating shareable links, or looking up vault contents.
using-docker-in-web
Set up and use Docker in Claude Code for Web environments. Use when Docker builds fail with bridge/iptables errors, SSL certificate errors through proxy, or when working in containerized web environments.
using-jina-reader
Fetch web page content using Jina Reader as a fallback when websites block direct access. Use when WebFetch returns 403, 406, captcha pages, empty content, Cloudflare blocks, or anti-bot responses. Also use when the user explicitly asks to use Jina Reader.
writing-svelte5
Svelte 5 syntax reference. Use when writing ANY Svelte component. Svelte 5 uses runes ($state, $derived, $effect, $props) instead of Svelte 4 patterns. Training data is heavily Svelte 4—this skill prevents outdated syntax.
maintaining-claude-code
Audit and improve Claude Code hooks, rules, and settings.json. Use when adding/debugging a hook, organizing rules, auditing settings.json (permissions, env vars, stale flags), or deciding between hook vs skill vs rule vs CLAUDE.md. For SKILL.md authoring use the skill-creator plugin; for CLAUDE.md audits use the claude-md-improver plugin.
managing-dotfiles
Manage dotfiles tracked in ~/.claude/dotfiles/ using GNU Stow. Use when adding config files to dotfiles, stowing/unstowing packages, or troubleshooting symlink conflicts.
cleaning-commit-history
Reorganize and clean up messy commit history on a feature branch into logical, reviewable commits. Use when cleaning up commits, preparing a branch for review, separating formatting from logic changes, fixing broken intermediate states, or squashing WIP commits.
committing-changes
Stage, validate, and commit changes with a clear message, optionally pushing to remote and monitoring CI. Use when committing code, creating a commit, pushing changes, or doing a commit-and-push workflow.
executing-test-plans
Execute a structured test plan and emit a concise results document (JSON + Markdown). Use when running test plans, executing QA workflows, verifying acceptance criteria systematically, or producing test results documentation.
cleaning-commit-history
Reorganize and clean up messy commit history on a feature branch into logical, reviewable commits. Use when cleaning up commits, preparing a branch for review, separating formatting from logic changes, fixing broken intermediate states, or squashing WIP commits.
committing-changes
Stage, validate, and commit changes with a clear message, optionally pushing to remote and monitoring CI. Use when committing code, creating a commit, pushing changes, or doing a commit-and-push workflow.
executing-test-plans
Execute a structured test plan and emit a concise results document (JSON + Markdown). Use when running test plans, executing QA workflows, verifying acceptance criteria systematically, or producing test results documentation.
monitoring-ci
Monitor and watch GitHub Actions CI/CD pipeline runs in real time. Use when the user has pushed code and wants to watch the build, check CI status, see if tests passed, monitor a pipeline, or wait for a workflow run to complete. Also use after any `jj git push` or `git push` command. Reports pass/fail with failed job logs. Do NOT use for writing or editing CI workflow YAML files, optimizing CI config, or debugging CI configuration — only for monitoring active runs.
searching-history
Search past Claude Code sessions to recall prior solutions, commands, fixes, and decisions. Use when the user references earlier work ("how did I fix", "last time we", "pick up where we left off", "a few months ago", "what was that … I set up", "didn't I once").
specifying-and-planning
Plan project implementation -- extract requirements from specs, break into tasks, create GitHub issues, set up project boards, or generate agent team configurations. Use when planning a project, extracting requirements, breaking down tasks, creating issues from a spec, setting up GitHub projects, or generating team plans.
syncing-claude-config
Sync this Claude Code config against new release notes — find config features added since the pinned baseline that this config should adopt. Use when Claude Code has been updated, when the user asks what's new / what changed / whether to adopt new settings or hooks, after a version bump, or when checking config currency against the latest release. Reports adoptable settings/hooks/env-vars/permissions as a risk-tiered proposal, then bumps the baseline. Do NOT use for auditing existing config quality, debugging hooks, or organizing rules — that is `maintaining-claude-code`.
using-jj
Jujutsu (jj) version control reference. MUST load this skill whenever a query mentions jj, jujutsu, revsets, absorb, evolog, oplog, immutable_heads, divergent changes, or jj-specific concepts — jj differs from git in non-obvious ways and the skill contains critical constraints (e.g., always pass -m, never use split/diffedit) that prevent broken workflows. Skip only for trivial jj commands you're certain about (describe, new, commit, push).
using-linear
Query, create, and manage Linear issues from the command line. Use when checking assigned work, viewing issue details, creating new issues, or updating issue status.
validating-project
Auto-detect and run formatters, linters, type checkers, and tests for the current project. Use when validating a project, running all checks, checking code quality before committing, or verifying the build passes.