parallel-agents
Pass
Audited by Gen Agent Trust Hub on Apr 24, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill is purely documentation-based, providing templates for agent interactions without any accompanying executable code or scripts.
- [NO_CODE]: No third-party dependencies, Python packages, or Node.js modules are referenced or installed.
- [DATA_EXFILTRATION]: No network operations or credential access patterns were found; the skill is limited to read-only local file system tools.
- [PROMPT_INJECTION]: The instructions do not contain attempts to override agent safety constraints or extract system prompts.
- [SAFE]: The skill implements a synthesis protocol that aggregates output from multiple agents. While this creates a surface for indirect instructions to enter the synthesis context, the risk is mitigated by the lack of automated command execution and the restricted toolset (Read, Glob, Grep). Ingestion points: Agent findings in SKILL.md synthesis protocol. Boundary markers: Markdown headers and tables. Capability inventory: Read, Glob, Grep, and Agent Tool. Sanitization: None.
Audit Metadata