prose-writer
Pass
Audited by Gen Agent Trust Hub on Mar 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns or security bypasses detected in instructions or scripts.
- [SAFE]: Implements a Human-in-the-Loop (HITL) approval mechanism, requiring the user to tick 'Approve C2' in 'DECISIONS.md' before prose writing is permitted.
- [SAFE]: Performs automated quality gating on inputs, verifying that briefs and evidence packs (subsection_briefs.jsonl, evidence_drafts.jsonl) are fully populated and free of placeholder markers.
- [SAFE]: Employs safe citation practices, restricting the agent to a local verified bibliography (ref.bib) and prohibiting the invention of sources.
- [SAFE]: Internal script 'run.py' uses standard path resolution to import local framework tools from the repository root, with no external network requests or unauthorized file access.
Audit Metadata