supply-chain-security
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection Surface: The skill is designed to perform security audits on software repositories, which requires the agent to ingest and analyze untrusted content such as workflow files, scripts, and configuration data. This process naturally creates a surface where malicious instructions embedded in a target repository could attempt to influence the agent's assessment.
- Ingestion points: As described in
references/capabilities-inventory.md, the agent is instructed to "Search the repository for evidence of the capability" across various files and documentation. - Boundary markers: The skill documentation does not currently specify the use of delimiters or specific instructions to the agent to ignore potentially malicious directives found within the files being analyzed.
- Capability inventory: The agent's task involves performing extensive read operations and directory listings on the target repository to evaluate its security posture.
- Sanitization: The reference materials do not define specific sanitization or validation protocols for the content retrieved from the repositories under review.
- Reference-Based Instructions: The skill includes specific "Critical incident prompts" within the reference data (e.g., in
references/capabilities-inventory.md). These are designed to guide the agent in reflecting on security incidents during the assessment process and do not represent attempts to bypass safety filters or override general agent behavior.
Audit Metadata