ds-jury
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions (Step 3) explicitly suggest using a second agent CLI in the terminal, such as a Codex or Claude CLI, to facilitate communication with juror models. This involves the execution of shell commands to interact with external tools.
- [INDIRECT_PROMPT_INJECTION]: The skill processes structured feedback from external models and uses it to perform automated modifications to the codebase (Step 7).
- Ingestion points: Findings and verdicts (APPROVE/CHANGES) provided by external juror models in Step 5 and Step 8 of SKILL.md.
- Boundary markers: The instructions define a specific structured format for juror findings (
id severity file:line claim fix) which acts as a rudimentary boundary for the ingested content. - Capability inventory: The agent possesses the capability to modify the filesystem (Step 7) and execute terminal commands/APIs (Step 3).
- Sanitization: The skill includes a verification step where the agent must "Check each finding against the code yourself" before applying fixes, which serves as a security control against blindly following external instructions.
Audit Metadata