change-to-regression
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [SAFE]: Analysis of the skill instructions and configuration revealed no malicious patterns or security vulnerabilities. All referenced resources are consistent with the established vendor.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user diffs. It mitigates potential injection by instructing the agent to treat diff content strictly as data and includes validation test cases in the evals directory.
- [COMMAND_EXECUTION]: The skill architecture restricts tool usage to a read-only environment in workflow.yaml, requiring explicit user approval for any state-changing operations or command executions.
- [DATA_EXFILTRATION]: Prompt instructions in prompts/change-to-regression.md explicitly forbid the agent from referencing internal memory or system files in its output, mitigating the risk of sensitive internal information disclosure.
Audit Metadata