research

Pass

Audited by Gen Agent Trust Hub on May 4, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill adheres to a structured five-step workflow that prioritizes observation and context-gathering over decision-making. It restricts network activity to fetching URLs provided by the user or discovered within the codebase, which is consistent with its stated purpose as a research tool.\n- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it ingests and processes data from external sources and the project codebase.\n
  • Ingestion points: Untrusted data enters the agent context through external URLs (Step 3) and codebase scanning (Step 2).\n
  • Boundary markers: The skill does not define explicit delimiters or 'ignore instructions' markers to isolate external content from the synthesis logic.\n
  • Capability inventory: The agent has the capability to write to the local filesystem (Step 4) and perform network fetch operations (Step 3).\n
  • Sanitization: No specific sanitization or validation of external documentation or codebase comments is described before the content is incorporated into the research artifact.
Audit Metadata
Risk Level
SAFE
Analyzed
May 4, 2026, 05:53 PM