session-delivery-hygiene
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFE
Full Analysis
- Functional Hygiene Guidelines: The skill consists of instructional content designed to guide an AI agent in maintaining a clean and accurate workspace. It includes guidance on verifying Git states, PR destinations, and cleaning up temporary resources like processes or files created during a task. These instructions promote safe operational practices and help prevent accidental overwriting of user data or environment settings.
- External Instruction Reference: The skill references a parent directory instruction file (
../../copilot-instructions.md). While this creates a dependency on an external file, it is a standard practice for managing shared agent instructions within a repository and does not represent an inherent security risk in this context. - Data Ingestion Surface: The skill instructs the agent to read live PR states and project configuration files (e.g.,
project.assets.json). While these are external data sources that could theoretically contain indirect instructions, the skill's own instructions focus on using this data for validation and provenance rather than executing untrusted content. This aligns with standard development workflows.
Audit Metadata