agent-communication
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION] (SAFE): The instructions in SKILL.md define a specific persona and prioritize provided reference files. While it instructs the agent to 'Ignore generic approaches,' this is context-specific to the domain of agent communication and does not attempt to bypass global safety filters or extract system prompts.
- [DATA_EXFILTRATION] (SAFE): No hardcoded credentials, sensitive file paths, or network-enabled commands (e.g., curl, wget) are present in the files.
- [REMOTE_CODE_EXECUTION] (SAFE): The skill does not define external dependencies or attempt to download/execute remote scripts.
- [INDIRECT_PROMPT_INJECTION] (LOW):
- Ingestion points: The skill is designed to handle 'agent communication' and 'message passing' (SKILL.md).
- Boundary markers: No specific boundary markers for untrusted agent data are defined.
- Capability inventory: No active tools or subprocess capabilities are present in the skill files.
- Sanitization: The skill explicitly mitigates injection risks by requiring 'schema validation' (references/validations.md) and 'constrained, validated communication' (SKILL.md).
Audit Metadata