retro
Pass
Audited by Gen Agent Trust Hub on May 7, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The instructions direct the agent to use shell commands and file modification tools to analyze errors and implement fixes like lint rules or tests. This involves standard development workflows.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes untrusted data such as git history and CI logs. Ingestion points: CI logs, git history, and source code (SKILL.md Step 1). Boundary markers: Absent; no delimiters are specified for processed content. Capability inventory: High; includes shell execution and file write capabilities. Sanitization: Absent; no input validation or sanitization is instructed for analyzed data.
Audit Metadata