cross-platform-cloud-verification
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
oz-devCLI to discover available runners and repository environments. These commands are limited to discovery and do not involve executing untrusted input. - [INDIRECT_PROMPT_INJECTION]: The skill aggregates reports and evidence from child agents, which constitutes an attack surface for indirect prompt injection. A subverted child agent or malicious test output could theoretically attempt to influence the parent agent's context through crafted artifacts.
- Ingestion points: Child agent responses, including status, results, and evidence artifacts, are ingested by the parent agent in the aggregation step (SKILL.md, Step 8).
- Boundary markers: The skill uses structured text blocks for child reports, which helps separate child-provided data from parent instructions.
- Capability inventory: The skill can discover infrastructure metadata and launch further remote agents via
run_agents. - Sanitization: The workflow does not specify explicit sanitization or escaping of the aggregated evidence content before it is processed by the parent agent.
Audit Metadata