vercel-common-errors
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the Vercel CLI (vercel) and curl to interact with Vercel's infrastructure, retrieving deployment logs, inspection details, and environment variable metadata for diagnostic purposes.
- [INDIRECT_PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection by ingesting deployment and function logs which are external, potentially untrusted sources.
- Ingestion points: Data enters the context through vercel logs, vercel inspect, and Vercel API responses via curl (specified in SKILL.md).
- Boundary markers: No specific delimiters or warnings are used to isolate ingested log content from the agent's core instructions.
- Capability inventory: The skill has the capability to execute shell commands (vercel, curl, jq, grep) and read files.
- Sanitization: There is no evidence of sanitization or filtering applied to the external log data before it is processed by the agent.
Audit Metadata