cloudflare-security-audit
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted codebases as its primary function, which can contain hidden instructions intended to manipulate the agent's behavior during the audit process.- Ingestion points: The 'Target' codebase to be audited, established during the setup phase (SKILL.md).- Boundary markers: The skill methodology does not define explicit delimiters or instructions to ignore embedded directives within the codebase being analyzed (SKILL.md).- Capability inventory: The methodology describes actions including probing, exploitation, persistence, and data extraction, which represents a high-impact capability set if an injection occurs (SKILL.md).- Sanitization: No sanitization, filtering, or escaping of the ingested code is mentioned before the agent or its sub-agents process it (SKILL.md).
Audit Metadata