deep-research-surf
Pass
Audited by Gen Agent Trust Hub on May 5, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Potential for indirect prompt injection via external data sources.
- Ingestion points: Untrusted data is retrieved from the internet using the
mcp__surf__*suite of tools (search, crawl, GitHub, Reddit, X, and YouTube) as described inSKILL.md. - Boundary markers: The instructions lack delimiters (e.g., XML tags or code blocks) or specific warnings to ignore instructions found within retrieved content from external sources.
- Capability inventory: The skill utilizes the
Tasktool to spawn parallel subagents inSKILL.md(Stage 1 and Stage 2), which process the untrusted content. - Sanitization: No validation, escaping, or sanitization of external data is mentioned before processing or final report synthesis.
Audit Metadata