maintain-verification-skill
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from project-local feature files and source code, creating an attack surface where malicious content in those files could attempt to influence the agent.\n
- Ingestion points: Project-local verification skill files and source code (e.g.,
.agents/skills/verify-<app>/).\n - Boundary markers: The skill provides isolation by instructing that sub-agent briefs stand alone and child agents operate in a read-only, no-write mode.\n
- Capability inventory: Capabilities include sub-task delegation (
delegate_task), UI/simulator automation (preview_*,device_*), and creating pull requests (link_pull_request).\n - Sanitization: No specific sanitization or filtering of the input file content is mentioned.\n- [DYNAMIC_EXECUTION]: The skill involves the creation or modification of harness scripts which are intended to be executable.\n
- Step 5 instructions direct the agent to fix harness gaps by ensuring scripts are executable and their invocation is documented, representing standard local script management for verification tasks.
Audit Metadata