auriga-go
Pass
Audited by Gen Agent Trust Hub on Apr 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection vulnerability surface. The skill relies on external repository files such as CLAUDE.md, task_plan.md, and GitHub pull request bodies to determine workflow steps, which could be exploited if these files contain malicious instructions. Ingestion points include processing workflow instructions from CLAUDE.md and PR descriptions retrieved via gh pr view. No explicit boundary markers or sanitization of instructions within these files are documented, and the skill is designed to drive tasks including file modifications, git operations, and test execution.
- [COMMAND_EXECUTION]: Local script execution and git/CLI tool interaction. The skill registers a bundled bash script scripts/ship-loop.sh as a Stop hook to manage its autonomous loop logic and directs the agent to utilize git and gh (GitHub CLI) for repository management tasks.
- [DATA_EXFILTRATION]: Access to session context. The ship-loop.sh script reads the agent's conversation transcript to detect completion signals, which involves processing the history of the current interaction locally.
Audit Metadata