subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill defines a methodical development process and contains no instructions or code aimed at compromising system security or bypassing safety controls. The core logic focuses on task management and quality assurance.
- [PROMPT_INJECTION]: The prompt templates provided (implementer, spec reviewer, and code quality reviewer) do not contain any instructions to override agent behavior or safety filters. The instructional emphasis on independent verification ('Do Not Trust the Report') is a standard methodology for ensuring task accuracy and does not constitute a malicious override.
- [DATA_EXFILTRATION]: There are no commands or instructions targeting sensitive system directories or attempting to transmit data to external domains. The skill operates within the context of the current development session.
- [REMOTE_CODE_EXECUTION]: The skill does not involve downloading scripts from external sources or performing remote executions. It coordinates standard development activities such as code implementation, testing, and git operations using internal tools.
Audit Metadata