Practical skills that teach AI agents to get things done for you. Built on the Agent Skills format.
npx skills add apocohq/skillsAlso works directly in Claude Code, OpenCode, and other AI tools that support the format.
Working in a large repo and losing track of where things live? This skill generates or updates distributed STRUCTURE.md files across directories, each with concise one-line summaries of folders and files. It helps LLMs orient quickly in unfamiliar codebases and is especially useful for knowledge repositories where discoverability and navigation matter.
Use when: mapping project structure, improving LLM repo orientation, indexing documentation-heavy repos, and keeping folder-level overviews up to date. Especially helpful when used as part of a pre-commit hook.
Tired of a messy inbox? This skill scans your Gmail, figures out who's emailing you, and organizes everything into clean Multiple Inbox sections. It generates a Google Apps Script that sets up labels and filters automatically. It can also help you find noisy senders worth unsubscribing from. It keeps a local config, so you can run it again to add new senders or tweak categories over time.
Use when: organizing Gmail, setting up multiple inboxes, managing labels and filters, cleaning up subscriptions.
Never sure what to focus on in the morning? Born out of spending 15–25 minutes every Monday manually sorting through 40+ todos (full story), this skill reviews your Things 3 todos, moves them into the right areas, tags what needs attention, and gives you a prioritized 30-second briefing so you know exactly where to start.
Features:
- Auto-generated config - On first run, scans your existing todos, areas, and tags to build a config with descriptions and examples. No manual setup needed.
- Smart categorization - Moves uncategorized items into the right area and applies tags conservatively using example-based matching.
- Daily routines - Automatically creates recurring weekday todos (e.g. "Check email") if they're missing from Today.
- Drucker-style briefing - Prioritizes your day into Must do / Should do / Could do using Peter Drucker's "Effective Executive" lens, with a motivational quote.
- Todo creation - Add todos by describing them naturally; the skill rephrases, assigns an area, and schedules them.
- Silent mode - Run with
[silent]for automated/headless execution via-pmode. Skips all prompts, requires existing config, and outputs only the final briefing. - Learning mode - Run with
[learning]to compare your todos against the config and refine area/tag descriptions and examples over time. - Idempotent - Safe to run multiple times a day. Skips items that already have areas and tags, and won't duplicate daily routine todos.
Use when: starting your day, triaging todos, organizing tasks by area and priority, adding new todos.
Have a pile of meeting recordings sitting in a folder? This skill watches meetings/incoming/ (and ~/Downloads/) for .vtt transcript files, parses them into clean Markdown with frontmatter, adds a title and summary, and files them into meetings/ with a date-slug filename. No manual reformatting needed.
Requires: uv installed on your machine (used to run the VTT parsing script).
Use when: processing meeting transcripts, converting VTT files to markdown, organizing meeting notes.
New repo? Inherited a project? This skill audits your .claude/settings.json and CLAUDE.md against team conventions and walks you through each gap one by one — attribution config, allowed tools, architecture principles, commit conventions, and recommended skills. It only applies changes you approve, so nothing gets overwritten by surprise.
Use when: bootstrapping Claude Code config in a new project, standardizing conventions across repos, or checking if a repo is aligned with team standards.
Got a meeting recording or transcript? This skill turns it into a clean, structured set of meeting minutes — decisions, action items, and key discussion points preserved verbatim. Works with .vtt files from Zoom, Teams, or any WebVTT-compatible recorder. Handles multilingual transcripts by translating to English automatically. Runs as a multi-stage pipeline to keep context lean even for hour-long meetings.
Requires: uv installed on your machine (used to run the VTT stripping script).
Use when: processing a meeting transcript, extracting action items, creating meeting notes, summarizing a recorded call.