review-api-design
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists entirely of instructional markdown and reference files. It does not include any scripts, binaries, or automated network operations. There are no patterns of obfuscation, privilege escalation, or persistence attempts.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided API designs through the
$ARGUMENTSvariable. An evaluation of the attack surface shows: - Ingestion points: Untrusted content enters the agent's context through the
$ARGUMENTSplaceholder inSKILL.md. - Boundary markers: The skill does not use specific delimiters or 'ignore' instructions to isolate the user-provided design from the skill's own instructions.
- Capability inventory: The skill is limited to generating a review document; it possesses no capabilities for writing files, making network requests, or executing system commands.
- Sanitization: No sanitization of the input design is performed. The risk is assessed as safe/negligible because the skill's primary instruction is to critically analyze and find faults in the provided design, which aligns the agent's behavior against following any instructions that might be embedded within that design.
Audit Metadata