implement-spec-with-codex
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides instructions for orchestrating code implementation using a delegation pattern. It emphasizes manual verification, integration, and review of all generated code, which are positive security practices.
- [INDIRECT_PROMPT_INJECTION]: While the skill processes external specifications and code diffs (ingestion points), which are potential surfaces for indirect prompt injection, it explicitly mandates that the agent performs thorough manual reviews and verification gates for all external output. This serves as a significant security control over untrusted data ingestion. The skill limits capabilities to local worktrees and git commits.
- [DATA_EXPOSURE_AND_EXFILTRATION]: No sensitive file paths, hardcoded credentials, or unauthorized network operations were detected. The skill operates on local worktrees and standard git development workflows.
- [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill does not perform any remote package installations or execute untrusted remote scripts. It uses internal delegation to other authorized skills within the platform context.
Audit Metadata