skills

Personal collection of agent skills, installable via the skills CLI.

Install

Option A — skills CLI (recommended; works with any agent)

Cross-agent install via skills.sh — Claude Code, Codex, Cursor, OpenCode, and more.

Install all skills globally:

npx skills add TimBroddin/skills --all

List available skills:

npx skills add TimBroddin/skills --list

Install one skill:

npx skills add TimBroddin/skills --skill youtube-research

Option B — Claude Code plugin (Claude Code only)

Installs the whole repo as a single plugin. All skills get registered at once.

/plugin install TimBroddin/skills

The plugin manifest lives at .claude-plugin/plugin.json — adding a new skill is one line there.

Skills

app-store-aso

Generate optimized Apple App Store metadata with ASO best practices, character-limit validation, competitive analysis, and screenshot strategy. Activates on App Store optimization, metadata review, or screenshot questions.

Pairs well with astro-mcp-server (full ASO MCP) and krankie (lightweight CLI for keyword rank tracking).

Includes a validate_metadata.py script that checks Apple's character limits independently of the agent.

codex-review-loop

Drive the GitHub Codex bot through repeated review-fix rounds on the current PR: post @codex review, wait for the bot to finish, auto-apply fixes for every actionable remark, push, and request another review — looping until the review comes back clean or a max-iteration cap is hit.

Detects this round's review precisely (anchors on the trigger timestamp, reads the bot's review + inline comments via the GitHub API) and waits adaptively — learning each round's duration instead of sleeping on a fixed guess. Treats remarks as reports to verify, not commands to obey, and reports any it deliberately skipped.

Requires gh (authenticated) and the Codex GitHub app installed on the repo.

pickup

Pick up an issue that's ready for an agent and work it end-to-end — from the project issue tracker to a pull request. The "Work" step of Matt Pocock's to-prd → to-issues → triage → pickup → ship pipeline.

Reads the tracker for issues carrying the AFK-ready triage label (default ready-for-agent), selects one (recommend-and-pick by default, or grab directly with /pickup <issue> / /pickup next), gates on whether it's genuinely ready, then delegates implementation to /tdd (enhancements) or /diagnose (bugs) and opens a PR. Issues that aren't ready get kicked back to /triage rather than half-built.

Requires Matt's skills installed and /setup-matt-pocock-skills run once per repo (defines the issue tracker and label vocabulary). Tracker-agnostic: works with GitHub, GitLab, or local-markdown issues per your config.

swift-missing-translations

Audit a Swift/SwiftUI project's Localizable.xcstrings (and AppShortcuts.xcstrings) for missing translations, compute per-language coverage, find raw source-language literals still hard-coded in UI code, and bulk-translate the gaps.

Reads the catalog's sourceLanguage and target locales from the project — works regardless of source language. Covers the trickier patterns the catalog alone can't fix: LocalizedStringKey vs String-typed surfaces, {name}-placeholder keys imported from JS i18n libraries, the Hashable-on-LocalizedStringResource compile error, the String(localized:) vs in-app-language-picker pitfall (with a LocaleBundle helper), and the AppShortcuts.xcstrings + AppEnum-for-Int-in-phrases setup for App Intents.

youtube-research

Deep LLM-driven research over one or more YouTube channels' videos. Lists each channel's catalog, filters videos by topic relevance, transcribes only the relevant ones, then synthesizes a single cross-channel research document with timestamped citations.

Subtitles-first via yt-dlp, with optional Whisper fallback. Workspace at ./.youtube-research/; final artifact in cwd.

Requires: yt-dlp, ffmpeg (brew install yt-dlp ffmpeg). Whisper optional.

Layout

skills/
└── <skill-name>/
    ├── SKILL.md          # frontmatter + instructions for the agent
    └── scripts/          # any helper scripts the skill calls

Each skill is self-contained — its SKILL.md declares its name and description, and any scripts live alongside it.

License

MIT