review-deep
Pass
Audited by Gen Agent Trust Hub on Jun 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell commands for environment cleanup (rm), directory management (mkdir), codebase bundling (repomix), and process monitoring (grep, tail, jq).
- [EXTERNAL_DOWNLOADS]: It uses
npx repomixto download and execute the Repomix utility from the NPM registry to prepare the codebase for analysis. This is a well-known developer tool. - [REMOTE_CODE_EXECUTION]: The skill orchestrates parallel sub-agents via the
opencodeplatform, executing dynamically constructed prompts across various Large Language Models. This is the intended primary function of the skill. - [PROMPT_INJECTION]: The orchestration logic includes instructions for the agents to operate 'headless' and 'non-interactively' without asking for user confirmation. While these resemble behavior-override patterns, they are appropriate for the tool's automated, batch-processing execution requirements.
- [DATA_EXFILTRATION]: The skill processes the local codebase for auditing but does not transmit data to unauthorized external domains. All communication occurs within the local environment or through the authorized orchestration platform.
Audit Metadata