security-audit
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to read and analyze untrusted project data (source code, lockfiles, manifests), which is a known vector for indirect prompt injection attacks.
- Ingestion points: Project files, dependency manifests, and lockfiles are ingested for analysis as specified in the audit protocol (SKILL.md).
- Boundary markers: The skill explicitly instructs the agent to treat all external input as hostile until validated, providing a clear instruction to maintain a security boundary.
- Capability inventory: The agent utilizes tool calls such as the
gitleakssecret scanner and the GitHub CLI (gh) for reporting findings, and it has the ability to delegate tasks to sub-agents. - Sanitization: While the protocol emphasizes identifying security vulnerabilities, it does not provide specific mechanisms for technical sanitization or escaping of the ingested code content before it is processed by the LLM, leaving a potential opening for prompt injection if the audited code contains malicious instructions.
Audit Metadata