Skills (this folder)

Heads up — there are two different “skills” concepts at PCT. Don’t confuse them.

What lives here (this folder)

Reusable agent workflow skills in Anthropic’s SKILL.md format. These are markdown-driven workflows that Claude Code (or Codex / Cowork) discovers and triggers on user intent. Each skill is its own subfolder:

skills/
  some-skill-name/
    SKILL.md
    references/
    assets/

The name and description fields in frontmatter are what tools read to decide when to trigger the skill. Be specific in description — vague descriptions cause wrong-trigger problems.

Use templates/skill.md as the starting point for new ones. This folder is currently empty — no general-purpose agent workflows have been authored yet for company-os.

What does NOT live here — the Brady-voice skills

The 12 production Brady-voice skills (brady-newsletter-intro, brady-ig-caption, brady-youtube-script, etc.) are a separate system. They are code-driven (Python + voice profile + Voyage retrieval + Claude API), not markdown-driven, and they live in their own repo for git versioning.

• Repo: bradyshearer/content-engine (private, on GitHub)
• OS overview: content-brain — what the Content Brain is and how to use it
• Operational reference: skills-architecture — where the files live across three layers, how Claude Code discovers them, setup on a new machine, naming conventions, adding new skills

Why kept separate: agent-workflow skills (this folder) describe what to do in markdown. The Brady-voice skills generate copy in Brady’s voice via Python that loads voice subprofiles, runs hybrid retrieval, and calls Claude with platform-specific prompts. Different runtimes, different inputs, different update paths. Mixing them would be confusing.

How to choose where a new skill goes

If the skill is…	File it in…
A markdown workflow that Claude Code reads from SKILL.md frontmatter to decide when to trigger	skills/{skill-name}/ (this folder)
Code-driven generation in Brady’s voice for a specific publishing channel	bradyshearer/content-engine repo (see skills-architecture)
Something Cloud-Functions-shaped (scheduled job, dashboard, sync, attribution)	A playbook entry in playbooks (e.g. attribution, content-brain)
A repeatable operational process that humans (or agents) execute	A playbook in playbooks