MylesMCook

Discover and use the current machine's nektos/act installation to validate or debug GitHub Actions locally. Use when act is requested, or when local execution of an existing CI test job would help authorized coding work. Check the installed CLI,…

Run a momentum-first three-lens Codex review of an explicit or readily inferable diff, commit, branch, plan, plugin, or small file set. Use when the user explicitly asks for adversarial, independent, three-reviewer, or Skeptic/Architect/Minimalist review.…

Create, plan, prompt, edit, and QA sparse 16:9 US English illustrations with Inkblot, a small solid-black deadpan worker. Use when the user asks for Inkblot or American Inkblot style, sparse white-background hand-drawn editorial/concept/article art, or a…

Build, integrate, debug, or update assistant-ui using current official upstream skills, documentation, and version-matched source. Use for assistant-ui questions, projects using @assistant-ui packages, or explicit assistant-ui evaluations. Generic chat UI…

Create, improve, format, or safely investigate evidence for fix-ready bug reports covering software, hardware, APIs, CLIs, SDKs, apps, services, integrations, data pipelines, infrastructure, build and test systems, and research tooling. Use when the requested…

Buzz CLI for relay operations: owner-reviewed agent drafts, messaging, channels, DMs, users, workflows, feed, reactions, canvas, social, repos, uploads, and agent memory.

Use this skill to write shell scripts or run shell commands to automate tasks in the browser or otherwise use Chrome DevTools via CLI.

Uses Chrome DevTools via MCP for efficient debugging, troubleshooting and browser automation. Use when debugging web pages, automating browser interactions, analyzing performance, or inspecting network requests. This skill does not apply to `--slim` mode (MCP…

Iterate on a running web app until a fresh first-time-user tester rates its usability 9/10. Each round a new, blind subagent tries real tasks in the browser and scores them; the main agent fixes what it reports and repeats. Use when the user asks for…

Inspect and safely archive stale idle Codex tasks when the user asks to clean up old Codex tasks, threads, or processes. Use native Codex task operations; do not delete tasks or create scheduled cleanup.

Use only when the user explicitly invokes $fugu-second-opinion or asks for Sakana Fugu or Fugu Ultra by name to provide a second opinion, external model check, tie-breaker, or sanity check. Calls the Sakana API with prompt-only, user-authorized context by…

Use when creating, planning, prompting, generating, or editing quirky US English article illustrations: blog, newsletter, essay, Notion, document body art, editorial explainer, workflow/methodology visual, concept metaphor, shot list, or sparse…

Use when the user wants to turn a rough idea, dirty goal, messy long-run prompt, sidechat thought, or under-scoped Codex task into clean Goal objective text under 4000 characters. Produce a ready-to-use Goal without the `/goal` prefix. Do not use for normal…

Use this skill when you need a serious code review, diff review, or implementation-plan review from independent reviewers. In Codex hosts, run exactly three fresh Codex subagents with Skeptic, Architect, and Minimalist lenses. Return a PASS, CONTESTED, or…

Use this skill when developing, debugging, refactoring, or reviewing Microsoft Excel Office Scripts (`ExcelScript` TypeScript) for workbook automation, Power Automate Run script actions, ranges, tables, charts, PivotTables, JSON, fetch, performance, or…

Use this as the default skill for MCP work: MCP servers, MCP Apps, ChatGPT apps, widget-backed connectors, and major MCP refactors. Follow one opinionated path by default: a hosted TypeScript `mcp-use` MCP Apps project with Streamable HTTP at `/mcp`,…

Use Brainerd as the umbrella router for repo-memory actions across Pi, Codex, and Claude. Repos that already have a brain may load Brainerd ambiently without invoking this skill.

Initialize a repo-local Brainerd brain for Claude, install the managed AGENTS.md block, write the thin CLAUDE.md shim, and sync Claude auto memory imports.

Persist durable learnings from the current Claude conversation into the repo-local brain using Claude session transcripts as the source of truth.

Mine older Claude session history for repeated corrections, preferences, missed durable knowledge, and recurring failure patterns in the current repo.

Initialize a repo-local brainerd brain for Codex. Use when the user asks to set up brainerd in a repo, create the repo brain, or install the managed AGENTS.md block that tells Codex to read the brain entrypoints.

Persist durable learnings from the current Codex conversation into the repo-local brain. Use when the user explicitly asks to reflect with brainerd, capture durable knowledge from this Codex thread, or update repo memory from this conversation.

Mine older Codex session history for repeated corrections, preferences, missed durable knowledge, and recurring failure patterns in the current repo. Use when the user explicitly asks to ruminate with brainerd, review older Codex repo history, or apply or…

Initialize a repo-local Brainerd brain for Pi. Use when the user explicitly invokes `/pi-init` or asks Pi to set up Brainerd in the current repo.

Create, refactor, and validate Pi packages. Use when an agent needs to scaffold a new Pi package, choose the correct Pi package surface, wire `package.json` `pi` manifests, or prove a Pi package works through the real `pi install` path.

Connect this installed Grasshopper plugin to a private server with the bundled client.

Exploratory browser testing for real web apps. Use when the user wants a smoke test, scoped verification, exploratory QA, or a balanced read on what works, what feels off, and what is blocked. Supports smoke, balanced, and bug-hunt modes. Uses agent-browser…

Reduce incidental complexity in code and design. Use when writing, reviewing, refactoring, or structuring code and when deciding API shape, module boundaries, layering, naming, error handling, testing strategy, or architecture. Prefer concrete, local,…