seo-ecommerce
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted product pages and marketplace data that could contain malicious instructions embedded in HTML or metadata.
- Ingestion points: Data is fetched from user-provided URLs via
render_page.pyanducp_check.pyas outlined inSKILL.md. - Boundary markers: The skill includes a dedicated 'Security' section in
SKILL.mdthat explicitly instructs the agent to treat fetched content as untrusted data and ignore any embedded commands. - Capability inventory: The skill possesses capabilities to render web pages, parse HTML, and interact with the DataForSEO Merchant API via local scripts.
- Sanitization: The instructions rely on the model's ability to follow safety guidelines rather than programmatic sanitization of the ingested content.
- [EXTERNAL_DOWNLOADS]: The skill performs automated network reachability probes based on endpoints defined in untrusted remote JSON files.
- Evidence: As described in
SKILL.md, theucp_check.pyscript uses a--probe-endpointsflag to send HEAD requests to URLs declared in a site's/.well-known/ucpprofile. - Risk: This mechanism represents a Server-Side Request Forgery (SSRF) surface if the remote configuration file points to internal or restricted network resources, although the skill documentation claims SSRF-blocked endpoints are reported.
Audit Metadata