Dan's Skills

Personal skills for AI coding agents. Many are fairly general-purpose — please enjoy.

(Designed for Claude Code but should work well with other AI agents that support the SKILL.md format.)

Skills

Skill	Description
CAB	Common Anchor Blueprint — structured folder systems with typed anchors, dispatch tables, and lint validation
Dev	Full development lifecycle — 9 stages from planning through release, with testing, debugging, and multi-agent orchestration
Ctrl	Local environment control — browser automation, persistent tmux sessions, system interaction
IO	External system I/O — Google Sheets/Slides/Drive, Dropbox, Notion, file sync via rclone
Edit	Visual editing — Excalidraw diagrams with SVG export and Obsidian embedding
MD	Markdown formatting — file trees, TOC, dispatch tables, cards, track changes
Product	Product research and purchasing — hunt, find, buy, reorder
Research	Investigation and synthesis — entity dossiers, topic surveys, book summaries
Role	Agent role definitions — pilot, pm, worker with persistent identity across sessions
Rule	Project rule management — define semantic rules, check code against them, triage exceptions, and drive fixes

Dev Skill — Development Lifecycle

The largest skill. 40+ actions organized into a numbered lifecycle:

Stage	Name	Actions
1x	Plan	anchor, prd, research, ux, system, plan-audit
2x	Architect	modules, system-design, test-plan, roadmap, arch-audit
3x	Implement	spec, code, test, review, verify, commit
4x	Release	changelog, version, package, publish, ship
	Capabilities
5x	Test	assess, scaffold, write-tests, verify
6x	Verify	rewire, lint, fix
7x	Adapt	ask-questions, replan
8x	Tactical	forge, debug, refactor
9x	Orchestrate	workers, worktrees, pr-flow, merge

CAB Skill — Common Anchor Blueprint

A system for organizing projects into typed anchor folders with consistent structure, dispatch tables for navigation, module documentation that tracks source code, and a lint tool that validates everything.

Action	What it does
/cab setup	Create a new anchor with full doc skeleton
/cab lint	Validate structure and module docs against source code
/cab tidy	Fix structural issues
/cab yore	Archive to Yore
/cab move	Rename/relocate an anchor
/cab migrate	Convert between anchor types

How Skills Work

Each skill lives in its own folder with a SKILL.md entry point. When a user types /skill-name action (e.g., /code plan, /research dig), the agent reads the corresponding action file and executes the workflow.

Skills are:

• Declarative — written in markdown, not code. The agent interprets the workflow instructions.
• Composable — skills reference each other (e.g., /code publish uses /md formatting conventions).
• Stateless — no runtime dependencies. Everything the agent needs is in the skill files.

Creating a New Skill

1. Create a folder under skills/ named after the skill
2. Add SKILL.md with YAML frontmatter (name, description, tools, user_invocable: true)
3. Add action files ({skill}-{action}.md) with workflow instructions
4. Register the skill in SKILL.md's dispatch table

---
name: myskill
description: >
  One-line description shown in the skills list.
  Use with an action argument: /myskill do-thing.
tools: Read, Write, Edit, Bash, Glob, Grep
user_invocable: true
---

License

MIT License. See LICENSE for details.