website-copy-reframe
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted content from external URLs or user-supplied drafts to generate scores and marketing copy.
- Ingestion points: Content is ingested from live URLs via scraping or from pasted drafts as described in
SKILL.md(Step 1). - Boundary markers: The instructions do not specify any delimiters or directives to ignore instructions found within the ingested data, which could allow embedded prompts to override the agent's intended behavior.
- Capability inventory: The skill can write files to the local system (markdown and .docx), interact with Google Docs, and perform network requests through scraping tools.
- Sanitization: There is no evidence of input sanitization or filtering to remove potential prompt injection vectors from the source material.
- [EXTERNAL_DOWNLOADS]: The skill instructions indicate the use of external web scraping tools (e.g., Firecrawl, Apify) to fetch content from live URLs for the audit and rewrite modes.
Audit Metadata