skills
Created: May 13, 2026
Last commit: September 11, 2026
Python64.3%
TypeScript14.2%
JavaScript12.8%
Shell4.7%
CSS1.5%
+6 moreREADME.md
skills
A collection of agent skills I use day-to-day with Claude Code, Codex, OpenCode, and other agent harnesses. Each skill is a self-contained folder with a SKILL.md (frontmatter + instructions) plus any supporting references, scripts, or templates.
Skills
Skills are grouped into category folders (review/, agent/, app-dev/, creative/, platform/, and others). Each skill is still a self-contained folder one level down (e.g. app-dev/chrome-cdp/), and the install loop below discovers them at any depth, so the runtime skill name is just the folder's basename regardless of which category it lives in.
The table below is regenerated by scripts/gen-readme-table.mjs. Re-run that script after adding or renaming a skill.
review
| Skill | What it does |
|---|---|
blast-radius | Find what a change could break somewhere else, then prove the one fact it's safe because of by running real code. |
conventions | Decide a code-style question from evidence in the repo rather than from preference: enumerate every existing instance of the construct, count the forms, and follow the dominant one (or the newest, if two coexist and the repo is mid-migration). |
docs-review | Review and update documentation against the current code. |
interrogate | Multi-reviewer adversarial review of a diff or change, with a counter-adversarial judge. |
perf-review | Performance-focused code review. |
repo-evolution-review | Analyze a repo (and optionally a sibling repo's commit history) and report two things separately: a narrative read of its theme, recurring patterns, and trajectory drawn from code plus commit history, and an objective good-and-bad patterns code-quality review judged against the repo's own AGENTS.md/CLAUDE.md conventions. |
robust-review | Maintainability-centered code review. |
testing-archive | Archive a verification/debugging campaign into a durable dated record in the repo docs — narrative README, curated screenshot/video evidence, re-runnable driver flows. |
update-review-ethos | Distill durable team preferences from a genuinely good MR review into the "Team code ethos" section of the team's shared agent-instructions file (CLAUDE.md). |
agent
| Skill | What it does |
|---|---|
c | Group the working tree's uncommitted changes into one or more logical commits, create them, then rebase onto the fetched default branch only if the branch no longer merges clean — whole files per commit, no per-hunk splitting, no push. |
clankie-mode | Sticky rigor router. |
clankie-perf | Clankie agent-orchestration performance hub (mission runs, lead/worker friction, skill efficacy) — archive orchestration runs as data points, aggregate metrics and trends, regenerate the static dashboard. |
co-w | Follow along with a sibling Herdr agent pane: watch its progress and offer targeted help. |
dir-rule | Add a directory-scoped rule to the nearest nested CLAUDE.md or AGENTS.md cleanly. |
herdr-handoff | Hand off a full conversation/work-session context to another agent pane in herdr: write a handoff file, pick or spawn the receiving pane, send a one-line kickoff prompt, verify pickup. |
herdr-lead | Lead an authorized Herdr swarm through concrete delivery: decide ready reviews, resolve ownership conflicts, dispatch bounded work and carry results through integration. |
linear-agent-session | Write an [Agent Session] handoff document in Linear at the end of a substantial agent work session, recording what a later agent or human can't re-derive from the code. |
linear-grind | Work the Linear queue to empty — pull every issue that is Todo or further along (Todo, In Progress, In Review), order it, then implement and close them one at a time until none are left. |
linear-issues | Write to Linear correctly and concisely — the API mechanics that silently destroy data (labels replace, relations append, pagination, media round-trip) and the editorial rules that size a post to the context a reader can't re-derive. |
linear-orient | Orient to a Linear project or issue before working or reporting its state. |
linear-plan | Turn the bounded work or plan in the current conversation into dependency-shaped Linear issues. |
linear-write | Voice overlay for writing Linear content — issue descriptions, acceptance criteria, comments, worklog notes, and design docs — in the user's own voice. |
mr-link | Commit, push, and hand back a create-MR/PR link plus a copy-paste command to run this worktree locally — then spin up an independent fresh-context reviewer on the pushed branch and adjudicate its findings. |
p | Push the current branch to the same-named branch on its remote, and nothing else — force-with-lease when a rebase requires it, never another branch, never a tag or a different remote. |
pr-description | Draft PR/MR descriptions by diffing the current branch against its default branch. |
pull-dev-repos | Pull/update all git repos under ~/dev (fast-forward only); optionally only third-party repos not owned by you. |
r-rule | Add a new rule to the current repo's agent instructions (CLAUDE.md / AGENTS.md) cleanly — integrating it into the right section, condensing/pruning for succinctness, and extracting oversized sections into their own skills. |
reflect | Mine the session just worked for durable lessons and propose where each keeper belongs — the skill that fired, a missed trigger, a new skill, a rule, or a mechanical check. |
rule | Add a new high-level agent rule to the current user-level AGENTS/CLAUDE instructions cleanly — first resolving this machine's source-of-truth file, then integrating, condensing, or extracting oversized procedure into skills. |
shared-checkout | Git discipline for committing in a checkout that other agents are working in at the same time. |
skill-maker | Create a new local agent skill using the user's symlink convention. |
solution-space | Before building glue, adapters, or workarounds on top of a system you don't own (cloud service, framework, library, third-party API), check what that system already models. |
work-tracking | How to use an issue/work tracker when one is actually in play — which tracker to use, how to assign work that needs the user, the lead/worker authorship model for multi-agent work on a shared tracker, and using the tracker as a diagnostic input when investigating bugs. |
app-dev
| Skill | What it does |
|---|---|
chrome-cdp | Inspect, measure, and screenshot a running Chrome tab programmatically via the Chrome DevTools Protocol. |
codebase-atlas | Scan a repository into a graph with the atlas CLI, and author the .codebase-index/_story.json narrative the Codebase Atlas story view reads. |
codebase-index | Build or incrementally update a repo's .codebase-index/ markdown mirror read by the nvim oil summary panel. |
excalidraw-skill | Programmatic canvas toolkit for creating, editing, and refining Excalidraw diagrams via MCP tools with real-time canvas sync. |
figjam-diagrams | Layout craft for readable FigJam architecture/flow diagrams. |
figma-design-systems | Apply a named visual system to FigJam boards through the official Figma MCP: palette, typography, panels, implementation-code panels, ERD tables, diagram nodes, and connector notation. |
framer-motion | Framer Motion / Motion for React patterns. |
frontend-builder | Design-to-code front-end builder. |
motion-from-reference | Reverse-engineer UI animation from a reference video (e.g. another app's screen recording) and reimplement it faithfully — ffmpeg frame extraction and montages, zero-dependency pixel measurement, spring/stagger parameter derivation, then re-record and verify with the same pipeline. |
obsidian-cli | Inspect, control, and edit a running Obsidian app/vault via the obsidian CLI (Chrome DevTools Protocol against the live app). |
obsidian-excalidraw | Create, edit, or debug Excalidraw diagrams inside an Obsidian vault. |
obsidian-vault-structure | Build or organize an Obsidian knowledge base — Map of Content (MOC) patterns, folder conventions, wikilink discipline, YAML frontmatter for note classification, graph-view-friendly shape. |
react | React and TSX guidance: state and effects, memoization and React Compiler, TanStack Query server state, TanStack Router, SSR/hydration/Suspense, composition, accessibility, testing, and animations. |
react-flow-v12 | Use for @xyflow/react (React Flow) v12+ work: node-based UIs, flow diagrams, DAG/workflow/canvas editors, custom nodes and edges, handles, layouts, viewport/state management, ReactFlowProvider, hooks, components, utilities, TypeScript types, SSR, color mode, migration from v11, and imports from '@xyflow/react'. |
react-native-skills | React Native and Expo best practices for building performant mobile apps. |
tldraw-design-systems | Apply a named design system to a tldraw board via document scripts: themes, panels, role nodes, ERD tables, sequence diagrams, toolbar tools, and connection meanings. |
creative
| Skill | What it does |
|---|---|
collage | Build moodboards and collages from a set of images using ImageMagick — scraping sources (e.g. tumblr), curating via contact sheets, themed color pools, and iterative render-look-fix layout. |
cut-talking-takes | Cut raw talking-head video (retakes of the same lines, OR a full walkthrough of a whole script) into a tight, modern jump-cut assembly — transcribe locally, drop dead footage/fumbles/false-starts/redundant re-explanations, keep the clean takes, and preserve 10-bit HDR color. |
davinci-resolve-scripting | Build and revise DaVinci Resolve projects and timelines through raw Python scripting. |
demoize | Drive polished desktop product demos: a deterministic driver script performs the API/browser/terminal beats while the human performs only what machines can't (record button, phone), gated by Enter cues. |
footage-index | Index a large footage, film, or gameplay-recording library so moments can be found by dialogue, visual content, or detected gameplay events (kills, multikills, hype), then imported into DaVinci Resolve without blanket transcoding. |
music-edit | Design, assemble, and refine music videos, fan edits, hype reels, and rhythmic supercuts from footage, a song, and optional style references. |
pokemon-rom-randomizer | Build deterministic, modded Pokémon ROMs headlessly with Universal Pokémon Randomizer ZX (Gen 1-4, GBA + NDS) and deliver them to Delta on iOS. |
psx-game-forge | Create and refine original PS1/PS2-era (and modern-nostalgia) 3D game assets, environments, characters, props, and the Blender-to-runtime seam (deterministic Blender Python generators, GLB export, verification in the real runtime). |
resolve-beat-fx | Apply and verify scripted effects, retiming, beat markers, and vertical reframing on a DaVinci Resolve timeline. |
song-stems | Separate a song into stems (vocals, bass, drums, keys...), analyze key/tempo/per-bar chords and the vocal melody, reharmonize with a new progression rendered as tempo-aligned MIDI, and remix the stems. |
tui-demo-gif | Record a terminal UI (TUI or CLI) as a polished demo GIF with vhs — it drives the real binary in a headless terminal from a scripted .tape — then verify the frames with ffmpeg. |
platform
| Skill | What it does |
|---|---|
figma-enhanced | Advanced Figma and FigJam editing through the official Figma MCP. |
local-llm-curator | Audit, maintain, and serve local LLM coding models on the user's Apple Silicon machine. |
md-share | Produce a standalone, shareable HTML snapshot of a markdown doc using md-preview's UI (pan/zoom mermaid diagrams, TOC, fullscreen). |
mini-shai-hulud-scan | Triage a macOS/Linux dev machine for compromise from the May 11, 2026 Mini Shai-Hulud npm + PyPI supply-chain attack (TanStack, Mistral AI, UiPath, OpenSearch, mistralai==2.4.6, guardrails-ai==0.10.1). |
morning-digest | Set up, operate, or modify a "morning digest" — a launchd job that runs headless Claude to summarize Slack (plus optional Linear/GitLab) and DM the summary to the user's own Slack DM, with a SwiftBar menu bar UI. |
nvim-config | Read the user's Neovim configuration before answering. |
record-unreal | Record real video of an Unreal Engine window on macOS — an editor viewport playing an animation, a PIE session, or a packaged game — for review or evidence, and know which capture method survives a shared desktop. |
resume-mgmt | Edit, rebuild, tailor, or publish James's resume — the HTML source in the portfolio repo, the generated PDF, and the structured resume.json the site and chat agent read. |
sst-ops | sst (Ion) deploy and state operations: bridging aws-login credentials, diagnosing stale locks, targeted-deploy/refresh hazards, and repairing a corrupted Pulumi snapshot. |
system-diagnostics | Diagnose local macOS CPU, GPU, thermal, process, stuck load-average, and fan issues. |
workstation-compromise-scan | General read-only compromise triage for macOS/Linux developer workstations, CI runners, and dev VMs: recent npm/PyPI supply-chain waves, poisoned AI-tool/MCP configs, GitHub workflow backdoors, suspicious persistence (launch agents, systemd units, cron), and CISA KEV exposure. |
x-video-grab | Download videos from X (Twitter) posts as the user scrolls, by attaching to a debug-mode Chrome tab and feeding permalinks to yt-dlp. |
principles
| Skill | What it does |
|---|---|
principle-encode-lessons-in-structure | Apply when writing the same instruction twice. |
principle-fix-root-causes | Apply when debugging. |
principle-guard-the-context-window | Apply when context is filling up. |
principle-model-the-domain | Apply to stateful or branchy code. |
principle-never-block-on-the-human | Apply when tempted to ask "should I?" on reversible work. |
principle-prove-it-works | Apply before declaring done. |
principle-redesign-from-first-principles | Apply when integrating a new requirement into an existing design. |
principle-sequence-verifiable-units | Apply to multi-step work. |
principle-subtract-before-you-add | Apply before an addition or refactor. |
root
| Skill | What it does |
|---|---|
md-preview | Preview local Markdown files or directories in a polished browser UI with live reload, file navigation, table of contents, syntax highlighting, and interactive Mermaid diagrams. |
Install
Clone the repo anywhere, then symlink each skill into the runtime roots your
agents read. The clone path does not matter — skill-maker auto-detects the source
root from its own location, and the loop below works off whatever REPO you
set. The local runtimes read ~/.claude/skills/, ~/.agents/skills/, and
~/.codex/skills/; all three resolve to the same source.
REPO="$HOME/dev/skills" # any path — skill-maker auto-detects its own location
git clone git@github.com:Volpestyle/skills.git "$REPO"
mkdir -p ~/.claude/skills ~/.agents/skills ~/.codex/skills
# Skills live under category folders, so discover them by their SKILL.md
# (works at any depth) and symlink each by its folder basename.
find "$REPO" -name SKILL.md -not -path '*/.git/*' | while read -r f; do
s="$(dirname "$f")"
name="$(basename "$s")"
ln -sfn "$s" ~/.claude/skills/"$name"
ln -sfn "$s" ~/.agents/skills/"$name"
ln -sfn "$s" ~/.codex/skills/"$name"
done
Some skills have additional setup (a running MCP server, a Chrome instance on a debug port, etc.) — each SKILL.md documents its prerequisites.
Authoring more skills
Use the skill-maker skill (or copy its scaffolding) to add a new skill. The convention enforced is:
- Source of truth lives in this repo (or another versioned repo).
~/.claude/skills/<name>,~/.agents/skills/<name>, and~/.codex/skills/<name>are always symlinks, never plain directories.SKILL.mdfrontmatter includesname,description, and — for skills that should auto-trigger on keywords rather than be invoked by name —user-invocable: false.
License
MIT — see LICENSE.