The Agent Skills Directory

[SAFE]: The skill contains structural instructions for managing agent teams. No malicious code, obfuscation, or unauthorized access patterns were identified in the orchestration logic.
[INDIRECT_PROMPT_INJECTION]: The skill defines a surface for ingesting planning documents to automate task creation, which is a standard operational pattern for orchestration.
Ingestion points: Processes prompt_plan.md and spec.md to determine team composition and task lists (SKILL.md).
Boundary markers: Implements a 'Plan Approval Mode' requiring manual authorization before any implementation changes are made.
Capability inventory: Utilizes SendMessage for spawning sub-agents and TaskCreate/TaskUpdate for task management.
Sanitization: Employs domain-based file ownership and explicit verification phases to ensure tasks are isolated and reviewed.

team-orchestrator