screenshots
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions guide the agent to construct and run shell commands, such as
gh pr diff <n> --name-only, by directly using values derived from user-supplied arguments ($ARGUMENTS). This logic creates a potential for command injection if the agent fails to sanitize these arguments before executing the shell command. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources within the repository environment and uses it to drive critical logic, such as determining which screens to capture. Ingestion points: Output from
git diffandgh pr diff, contents of router and pages directories, and user-provided arguments. Boundary markers: Absent; there are no instructions provided to the agent to treat this external data as untrusted or to isolate it from the instruction set. Capability inventory: File system access, UI screenshotting viaagent-browser, and pull request modification via theghCLI. Sanitization: Absent; the instructions do not specify any validation, filtering, or escaping for data retrieved from the environment.
Audit Metadata