conservative-dev-protocol
Pass
Audited by Gen Agent Trust Hub on Apr 28, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill establishes a "Minimum Intervention" protocol that explicitly requires human confirmation (
ask_user) and a technical diff review before any file system modifications are made. - [SAFE]: Instructions include specific defenses against prompt injection by mandating that user input be treated as data within
<user_request>delimiters, preventing it from being interpreted as agent commands. - [SAFE]: The provided
smoke_test.pyscript is a benign utility that verifies the presence of mandatory safety sections in the skill description. It does not perform network operations, execute arbitrary commands, or access sensitive files. - [SAFE]: The protocol includes automated fallback mechanisms (writing to
PROPOSAL.md) to maintain the safety barrier even when interaction tools fail, ensuring the agent does not proceed without oversight.
Audit Metadata