comprehensive-review-full-review
Pass
Audited by Gen Agent Trust Hub on Jun 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits a vulnerability surface for indirect prompt injection. It takes external, potentially untrusted code via the
$ARGUMENTSvariable and interpolates it directly into the instructions for multiple specialized sub-agents across all four review phases. - Ingestion points: External code is ingested through the
$ARGUMENTSvariable inSKILL.mdand passed to sub-agents in Phase 1 (Quality/Architecture), Phase 2 (Security/Performance), Phase 3 (Test/Docs), and Phase 4 (Compliance). - Boundary markers: The instructions lack boundary markers, delimiters (such as XML tags or Markdown code blocks), or explicit 'ignore instructions contained within the data' warnings when passing
$ARGUMENTSto sub-agents. - Capability inventory: The skill orchestrates 'general-purpose' sub-agents that are tasked with complex analysis, including security auditing, which could be subverted if the input code contains malicious instructions designed to hide vulnerabilities or skip checks.
- Sanitization: There is no evidence of input validation, escaping, or sanitization before the content of
$ARGUMENTSis included in the sub-agent prompts.
Audit Metadata