agent-skill-mapping
Skillの割当表(agent -> skills)を作るときのルール集。skill-assigner が mapping を更新するときに参照する。
When & Why to Use This Skill
This Claude skill provides a robust framework and rule set for managing agent-to-skill mappings within a multi-agent ecosystem. It optimizes orchestration by defining how capabilities are distributed, ensuring minimal redundancy, managing technical dependencies, and enforcing safety protocols for high-risk operations. It serves as the foundational logic for the skill-assigner to maintain accurate activation rules and system-wide mapping files.
Use Cases
- Agent Orchestration Optimization: Establishing clear rules to prevent overlapping skills across multiple agents, ensuring a lean and efficient multi-agent architecture.
- Dependency & Quality Control: Automatically ensuring that implementation-heavy skills are always paired with necessary verification tools like linters and testing suites.
- Risk Management & Safety Guardrails: Isolating dangerous or destructive operations into specialized agents that require human-in-the-loop review before execution.
- Dynamic Skill Mapping: Providing the source of truth for automated tools to update agent-skill-map.yaml and activation rules based on evolving project requirements.
| name | agent-skill-mapping |
|---|---|
| description | Skillの割当表(agent -> skills)を作るときのルール集。skill-assigner が mapping を更新するときに参照する。 |
Agent-Skill Mapping Rules
Inputs
.claude/agents/*.md.claude/skills/*/SKILL.md
Outputs
.claude/agent-skill-map.yaml.claude/agent-activation-rules.md
Rules
- 重複を減らす:同じ手順系Skillを複数agentへ持たせない
- 依存関係を明示:impl-* は検証(tests/lint)とセットにする
- 危険操作は隔離:必要なら専用agent + human review 必須
Verification
agent-skill-map.yamlがスキル名のタイポなしで参照していることactivation rulesに「どの入力でどのagentを起動するか」が書かれていること
Evidence
- N/A (bootstrap skill)