brida-reflex-page-clutter-relevance

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data extracted from web pages, creating a surface for indirect prompt injection attacks where malicious content on a site could influence the agent's behavior. * Ingestion points: The visibleText and goal fields in references/custom-reflex.json and references/playbook.md ingest content from the current browsing environment. * Boundary markers: The skill does not define specific prompt delimiters or markers to isolate the untrusted visibleText from the agent's instructions. * Capability inventory: While the skill is marked as recommendation-only in references/playbook.md, it is intended to guide agent decisions on element visibility within the browser context. * Sanitization: The playbook recommends 'deterministic safety/interaction signals computed by the host' and normalization, but lacks specific sanitization for the text content itself.
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to fetch availability and status information from external resources to verify capability availability. * The SKILL.md file directs the agent to read vendor-owned configuration files (/reflex/llms.txt, /reflex/agent.md, and /reflex/agent.json) from the vendor's live environment before making claims about feature support.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 08:22 PM
Security Audit — agent-trust-hub — brida-reflex-page-clutter-relevance