team-agent-orchestration
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill is entirely composed of markdown documentation defining an operating model for multi-agent orchestration. It contains no shell scripts, Python/Node.js code, or executable instructions.
- [INDIRECT_PROMPT_INJECTION]: The skill describes a workflow that ingests data from multiple agents, representing an indirect prompt injection surface at the orchestration level.
- Ingestion points: Work items, handoff artifacts (e.g.,
handoff.md), and agent-generated evidence stored in card metadata within SKILL.md. - Boundary markers: The instructions mandate the use of structured JSON schemas for work items and explicit "merge gates" to validate agent output.
- Capability inventory: The skill itself possesses no active capabilities (no file writes, network requests, or subprocess execution).
- Sanitization: The instructions focus on manual review and automated testing gates but do not specify automated sanitization for agent-to-agent communication.
Audit Metadata