ClawDiscover

PassAudited by ClawScan on May 1, 2026.

Overview

ClawDiscover appears to be a coherent instruction-only discovery skill, with disclosed external polling, optional paid endpoints, and optional webhook notifications that users should configure carefully.

This skill looks safe to install as documentation, but enable automation deliberately: keep polling on the free endpoint unless you approve paid x402 use, protect any webhook URL, and have the human review newly discovered services before the agent uses or installs them.

Findings (3)

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

Your agent could continue checking ClawDiscover periodically in the background if you add the heartbeat or cron configuration.

Why it was flagged

The skill recommends recurring autonomous polling. This is disclosed and purpose-aligned, but it means the agent may keep contacting the external service until the schedule is removed.

Skill content
Add to your heartbeat checks (every 6 hours recommended)
Recommendation

Only enable the schedule if you want recurring checks, keep the action limited to notification, and avoid automatic installation or use of discovered services.

What this means

If configured with x402 payment capability, the agent could incur small charges when using premium discovery, feed, or subscription endpoints.

Why it was flagged

The skill documents optional paid API endpoints. The costs are disclosed and no credential capture is shown, but using them may require payment authority.

Skill content
Paid Endpoints (x402) ... GET /api/v1/discover | $0.001 ... POST /api/v1/subscribe | $0.01
Recommendation

Set explicit spending limits or approval requirements for x402 requests, especially before combining paid endpoints with scheduled checks.

What this means

Your agent may expose a callback endpoint and receive external notifications that should not be treated as trusted instructions.

Why it was flagged

The subscription example shares an agent identifier, categories of interest, and a webhook URL with the external service, and it implies inbound webhook messages.

Skill content
POST /api/v1/subscribe ... {"agentId":"myagent","categories":["trading"],"webhookUrl":"https://myagent.com/webhook"}
Recommendation

Use a dedicated webhook URL with authentication or a secret, verify message origin, and treat webhook payloads as untrusted data for human review.