aeo-optimizer
Pass
Audited by Gen Agent Trust Hub on Jun 25, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill evaluates and transforms untrusted data, presenting a surface for indirect prompt injection. Ingestion points occur in SKILL.md Step 1 through user-pasted text or fetched URLs. Boundary markers are partially implemented via HTML comments used to delineate added capsules, but lack explicit instructions to ignore embedded commands in the source. Capability inventory reveals no access to subprocess calls, file-write operations, or privileged network actions beyond the initial fetch. No specific sanitization or filtering of the source text is described.
- [EXTERNAL_DOWNLOADS]: The skill uses network requests to fetch article content from external domains specified by the user to perform text analysis as a primary function of the tool.
Audit Metadata