skills

Mad House project lifecycle manager. Create, list, stage, promote, ship, and archive projects in ~/dev/mad-house/lab. Every project has a stage (concept → prototype → building → shipped → maintained → archived) and a type (code, game, creative, tool, experiment, content, bot). Use when managing lab projects, moving a project between stages, promoting from lab to a GitHub org, marking something as shipped, or archiving a project.

Bootstrap a new GitHub repo with all standard Mad House files. Creates the repo, clones it locally under the correct org path, and writes README, .gitignore, .env.example, Dockerfile, CHANGELOG, and AGENTS.md template. Triggers on "new repo", "bootstrap a repo", "create a new project", "start a new repo", "scaffold repo".

Manage the Mad House agent fleet. List all agents in the central registry, check which repos have which agents deployed, and sync agents from the registry to target repos. Use when adding a new agent to the fleet, syncing agents across repos, auditing what agents are deployed where, or rebuilding the agent registry. Triggers on "sync agents", "deploy agents", "what agents are deployed", "add agent to repo", "agent fleet", "update agents".

Generate SVG banners and navigation buttons for GitHub README files. Creates repeatable, customizable SVG assets - hero banners with title, subtitle, and tagline, and sets of rounded-corner navigation buttons with Inter font. Use when building or refreshing a repo README, updating branding, or generating any SVG visual assets. Triggers on "generate banner", "create svg banner", "make readme buttons", "banner for the readme", "design banner", "add banner to readme".

Run cargo fmt, clippy, and tests on a Rust workspace before committing. Catches format failures that break CI. Triggers on "cargo check", "check rust", "run rust tests", "before I commit rust", "fmt and test", "check the workspace", "is the rust build clean".

Check whether the last git commit in the current repo has the Claude co-author trailer. If missing, amend it. Triggers on "add co-author", "fix co-author", "missing co-author", "add claude to commit", "co-author trailer".

Surface your interpretation of the task before writing any code or making any changes. Use this at the START of any Build mode task - before opening a single file to edit. Critical when the request contained multiple ideas, was written in fragments, affects more than 2 files, or has any ambiguity about scope. Triggers on "build", "create", "add", "implement", "write", "make", "fix", "update" when the request is non-trivial. Do not skip this step to save time - a wrong interpretation wastes more time than a confirmation.

Deploy, restart, or check status of any application or service on the Coolify VPS. Lists all deployable resources, triggers deploys, tails logs, and confirms the container comes up healthy. Use when deploying a new version, restarting a service, checking deploy status, or reading deployment logs. Triggers on "deploy", "restart service", "push to production", "deploy to vps", "check deploy status", "deployment logs", "is it deployed".

Check whether the skill collection scripts are still accurate and up to date with the current environment. Detects drift between hardcoded lists in scripts and what actually exists - new AI tools installed, new projects added to ~/dev, new VPS containers, broken script paths. Use when the user says "sync my skills", "are my skills up to date", "update skill scripts", "check for drift", or after installing new tools or adding new projects. Also useful periodically to keep skills accurate.

Audit one or more GitHub orgs for stale live repos, local clones of archived repos, and repos missing standard files. Use when cleaning up the org, doing a monthly review, or when something feels disorganized. Triggers on "audit github", "audit the org", "org cleanup", "what repos are stale", "org health", "github audit".

Create a new Jupyter notebook (.ipynb) in the how-to repo. Use when the user wants to add a new guide, tutorial, or how-to. Triggers on "new notebook", "add a notebook", "create a how-to for", "add to how-to".

Audit the live Obsidian vault surface, plugin set, graph config, and drift between the canonical WSL vault and the live Windows vault. Use when the graph looks wrong, plugins misbehave, or vault changes are not appearing in Obsidian.

Improve Obsidian itself, including graph states, dashboards, plugin configuration, explorer order, file colors, Kanban, and other live vault UX surfaces. Use when the user wants Obsidian to feel better, not just the note content.

Mirror selected Obsidian surfaces between the canonical WSL vault and the live Windows vault using a safe allowlist. Use when graph settings, plugin data, or dashboards changed in one place and need to appear in the other immediately.

Full eagle-eye ops audit across both WSL and a remote VPS simultaneously. Use when the user wants a complete picture of everything - what's running locally and in production, what's live vs local-only, cross-environment dependency map. Triggers on phrases like "full audit", "audit everything", "ops audit", "what do I have running", "overview of my stack", "eagle eye view", or "show me everything".

Rename a local repo folder and update all AGENTS.md and memory references to the old name

End-of-session workspace sync - pull clean behind repos, push clean ahead repos, report dirty repos

Audit skills for quality, structure, and efficiency. Finds deterministic steps that should be scripts, duplicated logic across skills, composability gaps, and weak descriptions. Run after creating or updating any skill, or when the user says "audit my skills", "review my skills", "check skill quality", "is this skill good", or "improve my skills".

Batch add or remove frontmatter fields across one or more SKILL.md files

Package a locally built skill and publish it to the madebymadhouse/skills repo. Handles copying files, reviewing for internals to scrub, updating the README table, and pushing. Use when a skill has been built or updated locally and needs to go into the public skills repo. Triggers on "publish skill", "push this skill", "add skill to the repo", "package my skill".

Cleanly uninstall a CLI tool or package from WSL - finds every trace across npm, pip, cargo, config dirs, caches, VS Code extensions, and home files, then removes what the user confirms. Use when the user wants to fully remove a tool, package, or CLI from the WSL environment. Triggers on "uninstall", "remove", "get rid of", "clean up" followed by a tool name.

Deep audit of a remote VPS running Docker, Coolify, and Traefik. Use when the user wants to check what containers are running, inspect Coolify services, review Traefik routes, check disk/memory, review firewall rules, or get an eagle-eye view of their production server. Triggers on phrases like "audit the vps", "what's running on the server", "check production", "how's the vps doing", "what containers are up", or "check coolify".

Audit all git repos under ~/dev for dirty state, unpushed commits, exposed .env files, and tracked context files (CLAUDE.md, AGENTS.md, GEMINI.md). Use when the user wants to know if their workspace is clean, before a deploy, or when something feels off. Triggers on "audit workspace", "check repos", "anything unpushed", "is my workspace clean", "check for dirty repos", "any exposed secrets".

Automated drift detection across all repos and live services. Checks git repos for uncommitted changes and unpushed commits, pings service URLs for health, verifies VPS SSH connectivity, checks Coolify API status, and scans key npm projects for outdated packages. Outputs a machine-readable JSON summary and human-readable status. Triggers on "check for drift", "is everything healthy", "workspace drift", "check services", "are my services up", "what's drifted".

Find repos nested inside lab/, core/, prod/, or tooling/ and move them to a flat layout directly under ~/dev/mad-house/. Checks for unpushed commits before removing old locations. Triggers on "flatten workspace", "fix repo paths", "move repos to flat", "repos in wrong place", "clean up workspace structure".

Fast-forward pull all clean git repos under ~/dev. Skips dirty repos and repos with no remote. Safe to run at the start of any session to get all repos up to date. Triggers on "sync repos", "pull all repos", "update all repos", "get latest", "sync my workspace", "pull everything".

Deep audit of a WSL environment. Use when the user wants to understand what's running on their local machine, see the dev workspace, check runtimes, AI tooling, shell config, SSH keys, Docker state, or get an eagle-eye view of WSL. Triggers on phrases like "audit wsl", "what's on my machine", "check my local setup", "what projects do I have", or "show me my dev workspace".

Free disk space on WSL by clearing package manager caches (npm, pip, pnpm, playwright, node-gyp, puppeteer, prisma) and optionally pruning Docker build cache, images, and volumes on a remote VPS. Also reports lab node_modules. Triggers on "free disk space", "clean up caches", "WSL disk full", "debloat", "clear npm cache", "clean caches", "disk is getting low".

Replace with a description of what this skill does and when to use it. Include specific trigger phrases.

Cut a Rust crate release. Bumps version in Cargo.toml, updates CHANGELOG.md, runs full check suite, creates git tag, and pushes. Triggers on "release urchin", "bump version", "cut a release", "release X crate", "tag rust release", "ship rust version".

Take a visual screenshot of the Windows screen and read it. Claude can see and analyze whatever is currently displayed -- the Overseer TUI, a terminal, an error, anything. Use when asked to see the current state of the screen, debug a TUI, analyze what's displayed, or verify something visually.

Verify that AGENTS.md files are accurate by checking every path they reference against the actual filesystem. Reports stale paths, missing repos, and repos not mentioned in AGENTS.md. Triggers on "is agents.md accurate", "check agents.md", "sync agents.md", "verify agents.md", "agents.md drift", "update agents.md".

Add a new cell (markdown or code) to an existing Jupyter notebook in the how-to repo. Use when the user wants to extend an existing notebook with a new section, example, or code block. Triggers on "add a section to", "extend the notebook", "add cell to", "add content to the notebook".