vercel
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to utilize the
vercelCLI for operational tasks and thejr-rpcutility for configuration management. Evidence includes the use ofvercel logs,vercel inspect,vercel list, andjr-rpc config getcommands. - [INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by processing external data from Vercel logs while maintaining the ability to modify infrastructure state.
- Ingestion points: Build logs and project logs accessed via
vercel inspect --logsandvercel logsrespectively inSKILL.md. - Boundary markers: The skill provides behavioral boundaries by instructing the agent to 'Quote only decisive diagnostics' and warning that 'logs can contain private data'.
- Capability inventory: The skill can trigger Vercel deployments, assign aliases, and delete deployments as described in
SKILL.mdandSPEC.md. - Sanitization: No programmatic sanitization or escaping of log content is defined; the skill relies on the agent's instruction to be selective with output.
Audit Metadata