pal-e-docs

Skill Template

template-skill Template

active template
Project

Skill Template

Every skill note MUST have these sections. Skills define multi-step workflows with explicit MCP tool usage.

Template

Frontmatter Fields

Documents the YAML frontmatter in ~/.claude/skills/{name}/SKILL.md.
Field Purpose Values
<code>description</code> Shown in / autocomplete One-line summary
<code>context</code> Execution context <code>fork</code> = isolated subagent (no conversation history). Omit for inline execution.
<code>agent</code> Which subagent when context: fork Built-in (<code>Explore</code>, <code>Plan</code>, <code>general-purpose</code>) or custom from <code>.claude/agents/</code>
<code>disable-model-invocation</code> Restrict to user-only invocation <code>true</code> / omit
<code>argument-hint</code> Autocomplete hint for expected args E.g. <code>&lt;pr-number&gt;</code>
<code>user-invocable</code> Show in / menu <code>false</code> to hide / omit for visible
<code>allowed-tools</code> Tools permitted without asking Comma-separated tool names
<code>hooks</code> Skill-scoped hooks PreToolUse, PostToolUse, Stop
<code>model</code> Model override E.g. <code>haiku</code> for quick tasks

Wiring Pattern: context: fork + agent

Skills that delegate to agents use this pattern:
Important: context: fork + agent is convenience wiring, not enforcement. The forked agent inherits its own frontmatter restrictions, but true enforcement comes from hooks. See enforcement-architecture.

Naming Convention

Slug: skill-{name} (e.g., skill-implement-phase, skill-review-pr)
Tags: skill,active
Project: ai-agency

Skill Files

Each skill note should have a corresponding thin file in ~/.claude/skills/{name}/SKILL.md that points to the pal-e-docs note. The SKILL.md is the entry point (user-invokable via /name); the pal-e-docs note is the source of truth.
  • template-agent -- agent template (skills delegate to agents)
  • enforcement-architecture -- how skill wiring fits in the enforcement stack