pica-vercel-ai-sdk

Pass

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the use of npx @picahq/mcp to run the MCP server. This involves fetching and executing a package from the NPM registry. This package is a resource provided by the vendor (picahq).\n- [COMMAND_EXECUTION]: The documentation provides shell commands (find, grep) to be executed locally for diagnosing package version conflicts and searching documentation within node_modules.\n- [PROMPT_INJECTION]: The skill implements an MCP tool-calling interface, which is susceptible to indirect prompt injection if the data returned from third-party services (via Pica tools) contains adversarial instructions.\n
  • Ingestion points: Tool outputs from the @picahq/mcp client are integrated into the AI agent's context through the streamText function.\n
  • Boundary markers: No specific delimiters or "ignore instructions" guards are implemented in the provided code snippets to isolate tool results.\n
  • Capability inventory: The integration enables subprocess execution (via npx) and network access to integrated third-party services.\n
  • Sanitization: The skill relies on the default handling provided by the Vercel AI SDK and does not include custom sanitization of external data.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 28, 2026, 11:31 AM