manage-delegates
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill defines instructions for managing infrastructure using the harness-mcp-v2 server. All tools and resource types (harness_list, harness_create, delegate, delegate_token) are consistent with the vendor's provided infrastructure management capabilities. No evidence of credential theft, unauthorized data exfiltration, or remote code execution was found.
- [INDIRECT_PROMPT_INJECTION]: The process of listing and retrieving delegate/token statuses involves ingesting data that could theoretically contain hidden instructions, representing a standard injection surface.
- Ingestion points: The agent reads data returned by the harness_list and harness_get tools (SKILL.md).
- Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions to wrap external data.
- Capability inventory: The skill allows for the creation, modification (revoke), and deletion of delegate tokens via harness_create, harness_execute, and harness_delete.
- Sanitization: There is no documented logic for sanitizing or filtering data retrieved from the server before it is processed by the agent.
Audit Metadata