prune
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill performs read-only analysis of the agent's internal manifests, activity logs, and configuration profiles located in standard platform directories (e.g., ~/.claude/skills, .agents/PROJECT.md). This is consistent with its role as a portfolio cleanup specialist.
- [SAFE]: A 'propose-only' architectural constraint is enforced. The skill lacks the instructions or capability to execute file deletions or modifications directly, instead delegating these tasks to the user or an 'Architect' agent via structured proposals.
- [SAFE]: The skill includes explicit protections for a 'core' pack of essential skills (such as nexus, architect, and guardian), ensuring that critical agent functionality cannot be proposed for removal.
- [SAFE]: Recommendations are derived through a structured 5-axis retention scoring system (Usage, Overlap, Uniqueness, Coverage, and Maintenance). This mechanical approach reduces susceptibility to instruction-based manipulation.
- [SAFE]: No evidence of network exfiltration, credential harvesting, obfuscation, or unauthorized privilege escalation was found in the skill or its associated reference protocols.
Audit Metadata