research
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command 'gemini' with specific model and prompt flags if the binary is present in the environment. This represents local command execution with dynamically generated arguments based on research tasks.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources, creating a potential attack surface for indirect prompt injection.
- Ingestion points: Information is gathered from external websites via 'WebSearch' and GitHub repositories via 'docs-seeker'.
- Boundary markers: The instructions do not define boundary markers or provide explicit directions to the agent to ignore instructions embedded within the fetched research data.
- Capability inventory: The skill environment permits shell command execution ('gemini' bash command) and file system writes (saving reports to the 'Report:' path).
- Sanitization: There are no procedures specified for sanitizing, escaping, or validating the content retrieved from external sources before it is synthesized into the final report.
Audit Metadata