Airtable

AdvisoryAudited by Static analysis on Apr 30, 2026.

Overview

No suspicious patterns detected.

Findings (0)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

Installing the CLI gives local execution trust to whatever version npm resolves at install time.

Why it was flagged

The skill asks the user to install a global npm package at the latest version rather than a pinned version. This is normal for CLI setup but carries ordinary package provenance/version drift risk.

Skill content
npm install -g @membranehq/cli@latest
Recommendation

Install only from the expected npm package, consider pinning a reviewed version, and keep the CLI updated from trusted sources.

What this means

Once connected, the agent may be able to access Airtable resources allowed by the authenticated token.

Why it was flagged

The integration depends on delegated authenticated access to Airtable through Membrane, including token refresh. This is expected for the stated purpose but is sensitive account authority.

Skill content
Membrane handles authentication and credentials refresh automatically
Recommendation

Use the least-privileged Airtable/Membrane connection possible and revoke the connection when it is no longer needed.

What this means

A mistaken action or wrong record ID could update or delete Airtable records.

Why it was flagged

The documented Airtable actions include destructive and mutating operations. These are purpose-aligned for an Airtable management skill, but they can alter or delete live data.

Skill content
Delete Records | delete-records | Delete multiple records by their IDs (up to 10 at a time)
Recommendation

Require clear user confirmation for deletes or bulk updates, verify base/table/record IDs, and prefer read-only checks before mutation.

What this means

Raw API calls may bypass the clearer input schemas of named actions and can perform any Airtable API operation allowed by the token.

Why it was flagged

The skill provides a raw API proxy fallback that injects authentication headers. This is disclosed and useful, but broader than pre-defined scoped actions.

Skill content
send requests directly to the Airtable API through Membrane's proxy
Recommendation

Use named actions when possible, and review raw proxy requests carefully before sending them.

NoteMedium Confidence
ASI01: Agent Goal Hijack
What this means

Remote setup guidance could influence the agent's next steps if treated too broadly.

Why it was flagged

The workflow may expose provider-returned instructions to the agent. This appears intended for connection setup, but such instructions should not override the user's goal or safety checks.

Skill content
clientAction.agentInstructions (optional) — instructions for the AI agent on how to proceed programmatically.
Recommendation

Treat returned agent instructions as limited to the current Airtable/Membrane connection flow and ignore anything unrelated to the user's request.