deep-research

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection through its data ingestion and processing workflow.\n
  • Ingestion points: External content is retrieved using the WebFetch tool from arbitrary URLs in Step 2, and the skill reads from local source maps in Step 1.\n
  • Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded commands within the external content being processed.\n
  • Capability inventory: The agent is granted capabilities to write to the file system (Step 5) and generate new sub-tasks using the TaskCreate tool.\n
  • Sanitization: No sanitization or validation logic is specified for the data retrieved via WebFetch before it is used for extraction and evidence building.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 08:38 AM