ship-feature
Pass
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed with a robust security posture, utilizing a 'flight controller' pattern where a central agent manages state and delegating limited, bounded tasks to subagents. This architecture prevents subagents from making autonomous high-risk decisions without oversight.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies and mitigates the risk associated with processing untrusted external data (feature requests, GitHub issues, and repository source code) through a multi-stage review process.
- Ingestion points: Data enters the context from the feature request, repository instructions, issue context, and source code (SKILL.md, Section 1).
- Boundary markers: Subagent prompts are explicitly constrained by 'Common' and role-specific contracts defined in ROLE-CONTRACTS.md.
- Capability inventory: The skill possesses the ability to read/write files, execute shell commands (for testing), and perform Git/PR operations.
- Sanitization: Security is enforced via a mandatory 'SEC' quality lens and an independent 'Reviewer' role that actively challenges authorization, privacy, and data leakage claims.
- [COMMAND_EXECUTION]: Local command execution is restricted to the 'Implementer' role for the purpose of testing and validation. The skill mandates that the main agent inspect all resulting diffs and integration states to ensure that only intentional changes are committed, reducing the risk of malicious code injection.
Audit Metadata