commerce-detail-page-copy
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data through URLs or images when the user requests a page diagnosis. This data is processed to generate scores and recommendations, creating a vulnerability where instructions hidden in the analyzed content could influence the agent's behavior.
- Ingestion points: User-supplied URLs or screenshots provided to the
--mode diagnoseworkflow (SKILL.md). - Boundary markers: There are no explicit delimiters or instructions defined to isolate the external content from the prompt logic.
- Capability inventory: The skill generates structured JSON/Markdown copy and has the ability to chain to other specialized seller skills (SKILL.md).
- Sanitization: The skill incorporates an automated 'ai-slop-reviewer' step to filter repetitive or unnatural AI patterns, though this is not a dedicated security sandbox.
- [COMMAND_EXECUTION]: The skill attempts to verify the presence of a local 'codex' CLI tool (
codex --version) to use as an optional backend for its analysis logic.
Audit Metadata