pflow-task-finish
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFE
Full Analysis
- [CREDENTIALS_UNSAFE]: Configuration is loaded from a
.envfile in the repository root. The implementation uses a manual line-by-line parser that avoidssourceoreval, which prevents arbitrary code execution from the file content. - [INDIRECT_PROMPT_INJECTION]: The skill processes task titles that may contain untrusted data. It applies a dedicated escaping function to ensure this data is safely formatted for JSON output, preventing it from being misinterpreted as instructions by the agent.
- [DYNAMIC_EXECUTION]: The script sources a helper library from a related local skill (
pflow-commit). This dependency is managed using relative local paths and represents standard modularity within the author's toolset.
Audit Metadata