Claude Skills by jamesblackwell

Structured Codex, Claude, Amp, Pi, Kimi, or OpenCode code review when explicitly requested.

iOS Simulator automation using AXe CLI for touch gestures, text input, hardware buttons, screenshots, video recording, and accessibility inspection. Use when automating iOS Simulator interactions, writing UI tests, capturing screenshots/video, or inspecting accessibility elements. Triggers on iOS Simulator automation, AXe CLI usage, simulator tap/swipe/gesture commands, or accessibility testing tasks.

Read-only bug review for logic errors, race conditions, edge cases, runtime failures, and state bugs.

Use only when the user explicitly requests a deep investigation, deep dive, thorough investigation, or similarly exhaustive analysis. Do not invoke solely because a task is complex or research-heavy.

Cut, watch, and publish a signed notarized FalconDeck Mac GitHub Release (DMG). Use when asked to ship, release, tag, or publish the Mac/desktop app, a desktop-v* tag, a downloadable DMG, or GitHub Releases for desktop.

Check the diff against main and remove AI-generated slop introduced in the branch.

Manage secrets with Doppler: CLI operations, project/config/environment management, secrets injection, CI/CD integrations, and security best practices.

Reread recent code, catch obvious bugs/smells, and make focused cleanup fixes before handoff.

Find, evaluate, and report useful open-source GitHub reference implementations for a technical problem. Use when a user asks for existing code, open-source examples, prior art, a library comparison, or a proven implementation pattern before designing or building a feature.

Investigate rough objectives and turn them into clean, verifiable agent goals or task prompts for longer-horizon work.

Humanize text by removing AI-writing tells: hype, vague claims, formulaic rhythm, overused punctuation, and stock phrases.

Keep agents focused on requested code and tests. Use when work drifts into unnecessary audits, ceremonies, gates, or process.

Push a new FalconDeck mobile TestFlight build via EAS. Use when asked to push, ship, or release a TestFlight build.

Send James ntfy push notifications for completion, long-command results, blockers, or explicit notify requests.

Publish a substantial Markdown completion report as a secret GitHub gist and send James its link via ntfy. Use when James asks for a phone-friendly report link after long-running or detailed work, or explicitly invokes this skill; do not use for ordinary brief handoffs.

Extensive document-only UI QA spider. Inventory every reachable route or screen, then drive the real UI in a browser route-by-route and interaction-by-interaction, logging each bug with a screenshot. Use when the user asks for a full QA pass, QA spider, route-by-route audit, UX/UI audit, test every page, spider the app, exhaustive QA, mobile or desktop QA sweep, or runs /qa-spider. Do not use for a single-bug repro, a one-flow smoke test, or to implement fixes (this skill documents only).

Guide for writing idiomatic Rust code based on Apollo GraphQL best practices handbook.

Simplify and refine code. Only trigger on `/simplify-code`.

Create or update Codex skills with clear triggers, workflows, and tool usage guidance.

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

Run an extremely strict maintainability review for abstraction quality, giant files, and spaghetti-condition growth.

Condense verbose agent output or recent work into a 30-second executive summary — a single-line recap plus open decisions. Use when James asks for a TLDR.

React and Next.js performance optimization guidelines from Vercel Engineering for components, data fetching, and bundle size.

Run an explicitly requested long-running campaign to find and fix verified bugs across client and backend code, with regression evidence and Autoreview on every commit. Not for one-off debugging.

Use when controlling FalconDeck through its falcondeck MCP tools - creating or managing automations (scheduled agent instructions), changing agent-control settings, or inspecting runs and control history.

Drive a native GUI app (macOS, Windows, Linux) via the cua-driver CLI (default) or MCP server; snapshot its accessibility tree, act through snapshot-bound element tokens, native menu paths, exact window geometry, or pixel coordinates, and verify from fresh state. Use when the user asks you to operate, drive, automate, or perform a GUI task in a real application on the host, or to continue, resume, or recall recent Cua activity.

Use FalconDeck MCP tools during a turn. Call falcondeck_suggest_follow_ups near the end of work to offer next actions, falcondeck_rename_thread when the conversation's purpose has changed, and falcondeck_search/get/execute when the user wants automations or FalconDeck settings. Use these tools without waiting to be asked by name.