dev-build
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides a structured framework for incremental code implementation, focusing on testability and minimal implementation steps. It does not contain any malicious patterns such as credential exfiltration, persistence mechanisms, or unauthorized network operations.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from the user's repository, specifically 'repository instructions' and 'starting diffs' (Step 1). This creates a potential surface for indirect prompt injection if the project being analyzed contains malicious instructions.
- Ingestion points: Reads repository-level instructions and code diffs in Step 1.
- Boundary markers: No explicit boundary markers or isolation instructions are defined to separate repo data from the skill's logic.
- Capability inventory: The agent has the capability to write code, execute verification/test loops (Step 5), and perform cleanup/tidying tasks (Step 7).
- Sanitization: No specific sanitization or filtering of the repository data is implemented.
- [EXTERNAL_DOWNLOADS]: The skill includes a reference link to a GitHub repository hosted by Matt Pocock for attribution. This is a documentation reference only and does not involve automated downloads or execution of remote code.
Audit Metadata