survey
Pass
Audited by Gen Agent Trust Hub on Mar 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection. It uses the
WebFetchtool to ingest untrusted data from various external sources across four research lanes (Context, Solutions, Actual Behavior, and Alternatives). This data is then synthesized and written to the local filesystem using theWritetool. There are no safeguards described to prevent instructions embedded in the fetched content from being interpreted or executed by the agent. - Ingestion points: External web content retrieved via the
WebFetchtool during Step 1 of the workflow. - Boundary markers: Absent; the skill lacks instructions to wrap external data in delimiters or to use specific prompts to ignore embedded instructions.
- Capability inventory: The skill is authorized to use the
WriteandBashtools (specified in SKILL.md), which provide a significant capability set if the agent is influenced by malicious input. - Sanitization: No sanitization, validation, or filtering of the retrieved web data is implemented before it is processed and written to the disk.
Audit Metadata