Shelve every document into its proper home — CLAUDE.md, README, docs/, project skills, adr/, references/, notes/ — inspect each one's content against reality, and reorganize on approval or bootstrap the structure, constitution seed and the guarantees that close private memory, keep attribution accurate, and keep emoji out of files included, in a new project.
4 skills · 3 commands
Design and generate AI-only companies as repositories
3 skills · 3 commands
A spellbook of battle-tested Workflow scripts — cast them against a target, inscribe your own, and browse the collected spells.
1 skill · 3 commands
Conjure fenced autonomous routines that run on a schedule when you are away, each bound to report back so its work is never invisible.
1 skill · 3 commands
A personal eval harness that issues a report card to your Claude setup — register tasks as subjects, examine them with a fixed grader, and track how scores move when your configuration changes.
1 skill · 3 commands
Record the directions the user has rejected, and the principle behind each rejection, then replay them into every future session so a declined idea does not return in new clothes.
1 skill · 2 commands · hooks
A standard architecture for solo-developed products — a layered backend (Service Layer and Repository) under onion-style inward dependencies, thin entry points that all share one services layer, and views sorted by a single question — so Claude lays out a new project's skeleton and places each new file without the structure being explained again; scaffold the skeleton, and check an existing tree for violations.
1 skill · 2 commands
Catch the directional decisions made mid-conversation and record them as Architecture Decision Records, so future sessions read the reasoning instead of quietly reversing it.
1 skill · 2 commands
Turn your real Claude Code usage stats into an adventurer's log — unlock achievements from your actual sessions, tool calls, and late-night runs.
1 skill · 1 command · hooks
Translate an institutional document like your CLAUDE.md into the format and length limits of other AI services, so the same values carry across ChatGPT, Gemini, and generic system prompts.
1 skill · 2 commands
Distribute shared conventions, hooks, and settings fragments from an origin repository to a fleet of projects, and detect drift before it silently accumulates.
1 skill · 2 commands
Bottle up your session state, decisions, and next step so the next session picks up where you left off — a human-readable handoff, not a machine's internal state.
1 skill · 1 command · hooks
Call xAI Grok Build from the main Claude conversation like a subagent — MCP tools for task, resume, and doctor; sessions persist so corrections continue the same Grok session.
1 skill · 1 agent · 1 MCP server
Capture the terms you and Claude agree on into a project glossary and replay them into every future session, so a correction made once never has to be made again.
1 skill · 1 command · hooks
Delegate review feedback fixes to an agent team with file-based task splitting
1 skill · 2 commands
Capsule pasted content — reports, reviews, error logs from another session — with its source and date, so it survives the conversation and can be carried between sessions.
1 skill · 2 commands
Turn the conceptual explanations you were satisfied with into a Q&A notebook, so the next question in that area builds on the answer that already landed instead of starting over.
1 skill · 2 commands
Open the visual-verification loop: capture every visual surface of a project — simulators, web apps, videos, extensions, images — then look at the screenshots and judge them against intent instead of trusting that it works.
1 skill · 2 commands
When Claude's work is done but a human still has to verify it on real hardware or in production, emit a copy-pasteable handoff sheet so the baton pass never gets dropped.
1 skill · 1 command
Custom coding standards code review with local file output
1 skill · 1 command
Mine session transcripts for corrections you keep repeating and draft them into principles for your CLAUDE.md — you judge and ratify, the gardener only proposes.
1 skill · 1 command
Log which of your own skills and slash commands actually fire, then report what never did — so the plugin you built but that never triggers stops hiding.
1 command · hooks
Plant delightful easter eggs in your codebase
1 skill · 1 command
Dig your accumulated Claude usage history — command frequency, prompt shapes, usage rhythm, recurring manual work — and propose plugin candidates ranked by evidence, deduped against what you already have installed.
1 skill · 1 command
Post a session's actual outcome to Slack on stop, and brief you on stalled PRs, failing CI, and pending reviews across your repositories.
1 command · hooks
Draw a daily fortune for your Claude Code work — divined not from randomness but from your own recent usage data, once per day.
1 skill · 1 command
Spin up several independent attempts at one task in parallel, judge them through different lenses, and graft the winner with the best of the rest.
1 skill · 1 command
Watch your work for reusable structure and suggest turning it into a plugin, generated in an isolated background session
1 skill · hooks
Before a release, inspect whether the code change, the version bump, and the CHANGELOG entry are in sync across the repo and its packages, and report the mismatches — it checks, it does not fix.
1 skill · 1 command
Design a game's play as a choice between risk and return — no single move wins, clues are read rather than handed over, time pressure is not a reflex test — and measure every rule and number by playing the logic thousands of times without a screen, with the before and after side by side.
2 skills
Before offering concrete proposals for a vague goal, interview the user to draw out the picture in their head and agree on a one-page sketch of what to build.
1 skill · 1 command
Watch a session for waste-correlated behavior — repeated agent spawns and the same file read again and again — and quietly warn Claude before the tokens are gone.
1 command · hooks
Google Apps Script deployment management with clasp
1 skill
Help shape a verifiable completion condition for Claude Code's /goal command
1 skill
Keep the keys of an application's stored data: which fields to encrypt and what search, sort, and totals lose, where keys live and how they are backed up and rotated, what the operator holding the key can still see, and what an external API keeps of what you send it — checked against OWASP, NIST, and Japan's personal information guidelines.
1 skill
Product-agnostic principles for when and how a free app shows ads — never in front of the audience, whitelisted break points, two-stage gating, reward-driven quiet periods — distilled from a shipped iOS app's ADRs and checked against industry practice; Claude applies them when designing your next free tier.
1 skill
Scout Claude Code's latest release notes and translate each new feature into a concrete suggestion for your own setup and plugins.
1 command
One screen for every device, and every device checking the one screen. Lay a browser game's screen out at one reference size and scale the whole of it into any frame, the way an engine scales its canvas, with a foundation that re-measures safe-area insets an app's web view reports late, and a check that opens the game at real device sizes and reports every element that moves and every line of text that leaves its box.
1 skill
Turn image-generator "pixel art" into real pixel art a game can animate — find the generator's grid, read one colour per dot onto a shared palette, clear stray dots and doubled outlines — and check every frame against its original by overlap and colour difference, with a page that sets them side by side.
1 skill
Sprinkle life into AI-generated SVG — characters, spinners, icons, blobs, and illustrations alike: asymmetric organic shapes, layered tinted shadows, anticipation, squash and stretch, follow-through, and designed easing, so the output stops looking machine-made.
1 skill