airtable-cli
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches an installation script (install.sh) and standalone binaries from the official Airtable GitHub repository (Airtable/airtable-mcp-cli).
- [COMMAND_EXECUTION]: Executes the downloaded shell script to automate the setup process in the user's environment.
- [REMOTE_CODE_EXECUTION]: Downloads and installs platform-specific binaries from GitHub releases to provide a standalone CLI experience, utilizing checksum verification to ensure asset integrity.
- [DATA_EXFILTRATION]: Implements a security constraint in the URL validation logic that restricts the MCP server endpoint to the official airtable.com domain, preventing the exfiltration of personal access tokens to unauthorized servers.
- [PROMPT_INJECTION]: The skill ingests records and metadata from Airtable (via list-records, search-records, etc.), which presents an indirect prompt injection attack surface if the retrieved data contains malicious instructions.
- Ingestion points: Airtable records and table schemas read via the CLI (src/tools.ts) and processed by the agent (SKILL.md).
- Boundary markers: The CLI enforces structured JSON output, and instructions guide the agent to handle data as machine-parseable JSON.
- Capability inventory: The skill provides extensive write capabilities (create/update/delete records) via the MCP server interface (src/mcp.ts).
- Sanitization: The skill relies on the underlying Model Context Protocol (MCP) and official Airtable API/Server for data sanitization.
Audit Metadata