Skills
Navigate
skills/daangn/seed-design/seed-orchestrate-component/Gen Agent Trust Hub

seed-orchestrate-component

Pass

Audited by Gen Agent Trust Hub on Oct 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill uses a multi-agent orchestration pattern where a coordinator agent passes task instructions and data to worker agents, creating a potential surface for indirect prompt injection.\n
  • Ingestion points: references/collaboration.md indicates the ingestion of user requirements, existing component source files, and shared work memory documents.\n
  • Capability inventory: references/harnesses.md describes the use of orca, task, hub, and SendMessage tools to create and communicate with subagents.\n
  • Boundary markers: The skill establishes clear role boundaries and delegation protocols in references/roles.md and references/collaboration.md.\n
  • Sanitization: There is no explicit requirement for automated sanitization of data exchanged between agents, relying instead on role instructions.\n- [COMMAND_EXECUTION]: The skill documentation includes commands for interacting with the platform's orchestration runtime.\n
  • Evidence: references/harnesses.md contains status and capability checks for the Orca system, such as orca skills get orchestration and orca status --json.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 11, 2026, 10:15 PM