team-agent-orchestration

Pass

Audited by Gen Agent Trust Hub on Aug 13, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill establishes an orchestration workflow that ingests data from external agent outputs and files, creating a vulnerability surface where malicious instructions could be embedded in data.
  • Ingestion points: The agent is instructed to process "handoff artifacts" (e.g., handoff.md), "evidence" logs, "review notes", and "acceptance criteria" fields within JSON orchestration cards.
  • Boundary markers: The instructions do not specify any delimiters, escaping mechanisms, or explicit directives to ignore potential instructions embedded within handoff artifacts or external data.
  • Capability inventory: The orchestration process involves several high-privilege capabilities, including running agent squads ("Run agents"), merging code into mainline branches ("PR merged" / "local main updated"), and creating/modifying files to promote new skills ("promote it into skills/").
  • Sanitization: No sanitization, validation, or structural checking of ingested handoff content or evidence reports is described in the skill instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 13, 2026, 12:56 AM
Security Audit — agent-trust-hub — team-agent-orchestration