What problem does it solve? When a repository has dozens of skills, deciding which skill should handle a request becomes error-prone: requests span multiple skills, ownership is unclear, and unapproved or destructive steps can slip through. This Skill coordinates the whole workflow by clarifying intent, resolving the correct owning skill, enforcing policy gates, and tracking execution state. ## Core Features & Use Cases - Intent Routing and Ownership Enforcement: Resolves clarified user intent to exactly one owning skill per step using the skill inventory and ownership map, preventing overlapping claims. - Policy-Gated Execution: Blocks, denies, or requires explicit approval before dispatching destructive, external, privileged, or scope-expanding steps. - Workflow State Tracking: Emits completion artifacts (reports/workflow-completion.json and .md) summarizing executed, skipped, blocked, and pending steps with rationale. - Use Case: A user asks to "add a new linting skill and update the review workflow." The orchestrator routes the creation request to skill-create for duplicate analysis, routes the modification to skill-update after target confirmation, and halts any step lacking policy approval. ## Quick Start Ask the agent to coordinate a multi-step request, for example: route my request to add a new deployment skill and update the existing review skill through the proper owning skills.
