audit-auth-flows
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze application source code, which serves as an external, untrusted data source. This creates an attack surface for indirect prompt injection if the audited code contains instructions meant to influence the agent's behavior.
- Ingestion points: The agent reads project-specific files (e.g.,
middleware.ts, route handlers, server actions) and usesgrepto scan server code paths. - Boundary markers: The skill mandates a "Read-only" approach and explicitly warns "Do not run these against production."
- Capability inventory: The skill utilizes file system reading,
grepcommands for static analysis, and performs network probes (HTTP requests) during Phase 5 to verify authorization gates. - Sanitization: No automated sanitization or filtering of the ingested source code is specified in the instructions.
- [COMMAND_EXECUTION]: The skill instructs the agent to use
grepfor code discovery and perform live network probes in Phase 5. These are legitimate auditing actions within the scope of the skill's purpose and include instructions to use these capabilities non-destructively.
Audit Metadata