slice
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data such as project briefs, repository instructions, and existing codebases. While this creates a surface for indirect prompt injection, the instructions explicitly guide the agent to validate untrusted data at runtime and enforce server-side authorization. This indicates a defensive posture against common web vulnerabilities.
- [DATA_EXPOSURE_EXFILTRATION]: The skill includes specific warnings against logging sensitive payloads, secrets, or tokens. It recommends logging only identifiers and outcomes to prevent accidental data exposure in error logs.
- [PRIVILEGE_ESCALATION]: The instructions explicitly forbid the agent from changing account permissions, publishing, or merging code solely for the purpose of demonstrating task completion, maintaining a least-privilege approach during development.
- [COMMAND_EXECUTION]: While the skill involves implementing code changes, it does not include any direct or hidden commands that would execute malicious code on the host system. It focuses on repository checks and verification through isolated test data.
Audit Metadata