survey
Pass
Audited by Gen Agent Trust Hub on Apr 25, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill ingests untrusted external data through research activities, which presents an indirect prompt injection surface.\n
- Ingestion points: External web content retrieved via the WebFetch tool for landscape analysis (SKILL.md).\n
- Boundary markers: Absent. No delimiters or 'ignore' instructions are used to separate fetched content from instructions.\n
- Capability inventory: The skill utilizes Bash, Write, Read, and WebFetch tools.\n
- Sanitization: Absent. External research data is processed into markdown artifacts without validation or escaping.\n- [COMMAND_EXECUTION]: The skill executes a Python script provided by the author to validate the structure of generated survey artifacts.\n
- Evidence: The instruction to run 'python3 .agent-skills/survey/scripts/validate_survey_artifacts.py'.\n
- Context: This is an internal tool used for artifact consistency check during research tasks.\n- [EXTERNAL_DOWNLOADS]: The skill utilizes the WebFetch tool and provides instructions for programmatic retrieval of external evidence.\n
- Evidence: Use of the WebFetch tool and documentation in 'references/evidence-recovery-ladder.md' recommending 'python3 + urllib.request'.\n
- Context: These operations are core to the research purpose of the skill to gather information from external sources.
Audit Metadata