Argus Intelligence
PassAudited by ClawScan on May 10, 2026.
Overview
Argus appears to be a coherent instruction-only external API skill for blockchain risk checks, with user-notice items around sending data to ARGUS, optional payment tokens, A2A/webhooks, and release/provenance hygiene.
This skill looks purpose-aligned, but install it only if you are comfortable sending selected blockchain addresses, prompts, agent IDs, and webhook event data to ARGUS or the configured endpoint. Use the trusted endpoint, protect payment tokens, approve paid calls and webhook registration, and verify the full package/release because the provided SKILL.md view is truncated and file versions differ from registry metadata.
Findings (4)
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.
Blockchain addresses, prompts, agent IDs, or other query content provided for analysis may be sent to the configured ARGUS endpoint; paid endpoints may incur charges if payment credentials are used.
The skill's core workflow is sending curl requests to a user-configured external API endpoint. This is disclosed and central to the purpose, but it is still network/tool use that should be scoped.
curl -X POST $ARGUS_ENDPOINT/api/v1/free/query ... -d '{"query": "Is this address safe: 0x742d35Cc...", "agentId": "my-agent"}'Set ARGUS_ENDPOINT only to the trusted ARGUS URL, review data before sending, and require explicit approval before using paid endpoints.
If a payment token or proof is exposed or sent to the wrong endpoint, credits could be consumed or account/payment activity could be revealed.
The skill documents optional payment credentials/proofs for paid API access. This is expected for the service, but those values can authorize spending or credit use.
Buy 350 credits for $10 → receive a token → pass `X-Stripe-Token: <token>` header. ... Send USDC to treasury, attach `X-Payment-Proof` header.
Treat payment tokens as secrets, avoid pasting them into shared logs or chats, and only send them to the intended ARGUS endpoint after confirming the request.
Natural-language queries and event notifications may move between the user's agent, ARGUS, and a configured webhook URL; webhook events should be authenticated to prevent spoofing.
The skill explicitly supports agent-to-agent messages and webhook callbacks. These are purpose-aligned integrations, but they cross service and agent boundaries.
ARGUS supports the A2A protocol ... curl -X POST $ARGUS_ENDPOINT/message ... `POST /api/v1/webhooks/register` | Subscribe to real-time event alerts
Use HTTPS webhook URLs you control, validate the returned webhook secret/origin on incoming events, and avoid sending sensitive prompts or identifiers unless you are comfortable sharing them with ARGUS.
It may be harder to confirm exactly which release or source revision is being installed.
The supplied package files show version 1.9.2 while registry metadata says 1.9.3, and the registry source is listed as unknown despite homepage/repository fields. This is a provenance/version hygiene issue, not evidence of unsafe behavior.
Source: unknown; Registry metadata: Version: 1.9.3
Verify the publisher, repository, and full SKILL.md contents before relying on release-specific claims.