proposal-response
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFE
Full Analysis
- Authority Boundary Enforcement: The skill implements strict controls to prevent unauthorized data release. It explicitly enforces 'internal_review_only' statuses and refuses to mark content as approved, authorized, or released, ensuring that the final authority remains with human reviewers.
- Data Ingestion and Instruction Isolation: The instructions include a specific flow to treat supplied questions and attachments as data, explicitly directing the agent to ignore any embedded instructions within that data that might attempt to override the skill's defined workflow.
- Secure File Handling: The skill uses workspace-aware file operations and limits its persistent state to a dedicated tracking directory (
.copilot-tracking/proposal-responses/). It includes validation steps for existing artifacts to prevent overwriting malformed or unauthorized content. - Traceability and Evidence Mapping: By requiring every claim to cite a registered, read-only source artifact, the skill maintains high traceability and prevents the generation of unsupported or fabricated information.
Audit Metadata