research
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes a shell command using a CLI tool to perform research.\n
- Evidence: Phase 2 instructions state:
execute gemini -m gemini-2.5-flash -p "...your search prompt..." bash command.\n - The skill constructs this command by interpolating dynamically generated search prompts into the string, which may allow for command injection if input contains shell metacharacters.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize large amounts of data from external, untrusted sources.\n
- Ingestion points: The skill reads output from the
geminiCLI tool, theWebSearchtool, and GitHub repositories via thedocs-seekerskill.\n - Boundary markers: The instructions do not provide explicit boundary markers or guidance to ignore instructions embedded in the external content.\n
- Capability inventory: The agent has the capability to execute shell commands (
gemini) and write markdown files to the local file system (./plans/<plan-name>/reports/).\n - Sanitization: There are no requirements to sanitize or escape the content fetched from the web before it is processed by the agent.
Audit Metadata