audit-fe-api
Fail
Audited by Gen Agent Trust Hub on Jul 24, 2026
Risk Level: HIGHDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: Accesses sensitive configuration files.
- The skill instructs the agent to read
.envand.env.localfiles in Step 0d to extract port information. These files frequently contain sensitive secrets, credentials, and API keys that should not be accessed by general-purpose skills. - [PROMPT_INJECTION]: Indirect prompt injection surface via multi-source data ingestion.
- Ingestion points: The skill fetches external data from documentation (via Context7) and web searches (via Firecrawl) in Phase 1 to determine best practices.
- Boundary markers: Absent. The instructions do not define delimiters or provide warnings to the agent to ignore embedded instructions within the retrieved external content.
- Capability inventory: The skill possesses extensive read capabilities across the local file system (via Grep), production error monitoring (Sentry), and database schemas (Supabase). The skill description also mentions replacing code patterns, implying file-write capabilities.
- Sanitization: Absent. The skill does not validate or sanitize external content before using it to influence the agent's audit findings and code modification recommendations.
Recommendations
- AI detected serious security threats
Audit Metadata