dispatching-agent-teams
Pass
Audited by Gen Agent Trust Hub on May 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security issues were identified. The skill acts as a coordinator for agent teams using local files and platform-native tools.
- [PROMPT_INJECTION]: The instructions include positive constraints and coordination rules that prevent the agent from improvising workflows or skipping validation gates.
- [COMMAND_EXECUTION]: The orchestration process uses high-level agent tools for spawning and task management; no arbitrary shell or system command execution is present.
- [DATA_EXFILTRATION]: Data flow is localized to passing outputs between phases within the agent team, with no evidence of external exfiltration or unauthorized access to sensitive files.
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits a potential surface for indirect injection as it reads team definitions from local files. Ingestion points:
teams/{name}.mdfiles. Boundary markers: The skill relies on the file being a contract but does not implement specific delimiters for untrusted content. Capability inventory: Spawning sub-agents and preloading skills across peer, sequential, and inline modes. Sanitization: No explicit content validation or sanitization is mentioned, though the skill requires manual confirmation of entry and exit criteria.
Audit Metadata