Claude-skill-registry agent-invocation
Agent invocation syntax and boundary rules
install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/agent-invocation" ~/.claude/skills/majiayu000-claude-skill-registry-agent-invocation && rm -rf "$T"
manifest:
skills/data/agent-invocation/SKILL.mdsource content
Agent Invocation Rules
Invocation Syntax
Use
@agent-name followed by the task:
@api-researcher Research GitHub webhooks @api-architect Design the OpenAPI spec
Agent Boundaries
- Each agent has EXCLUSIVE responsibilities
- NO overlap between agents allowed
- Check agent descriptions for triggers
- Respect "does NOT" boundaries
Context Passing
When passing context between agents:
- Save output to memory folder
- Reference previous phase outputs
- Include all necessary context
- Use agent-parameter-passing skill
Agent Loading
Each agent must:
- Load its required rules
- Assert its responsibilities
- Check for prerequisites
- Validate inputs
Collaboration Rules
- Agents work sequentially, not in parallel
- Each agent completes before next starts
- Use memory files for handoffs
- Document decisions made