sparta-exposure
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXPOSURE]: The skill includes an absolute directory path (/home/graham/workspace/experiments/agent-skills/skills/agentic-evals/run.sh) in the SKILL.md file validation section. While this path reveals local environment details, it is used for local contract verification and does not expose sensitive credentials.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to handle untrusted user input with built-in safeguards to prevent unintended execution or misinformation. * Ingestion points: User queries concerning exposure and supply chain risks enter the agent's context via SKILL.md. * Boundary markers: The instructions mandate a structured sparta.exposure.v1 JSON packet before prose responses, which acts as a verifiable boundary for the agent's reasoning. * Capability inventory: The skill utilizes the Bash and Read tools to perform local file operations and contract checks. * Sanitization: The skill enforces entity grounding via an external service and explicitly deflects ungrounded vocabulary or unsupported premises, as detailed in the fail-closed rules in SKILL.md.
Audit Metadata