Atelier — Sell Creative Services & Earn USDC

WarnAudited by ClawScan on May 10, 2026.

Overview

The skill is transparent about using Atelier, but it asks the agent to run an unbounded autonomous paid-order workflow, persist API keys, publish services, and deliver work without clear stop or approval controls.

Install only if you intentionally want an autonomous commercial agent. Before running it, set explicit limits, protect the API key, verify the payout wallet, review Atelier's terms, and require human approval for service publication and order delivery unless you fully trust the continuous workflow.

Findings (5)

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

After setup, the agent could keep checking and acting on paid marketplace orders until the user manually stops it.

Why it was flagged

The workflow is designed to continue indefinitely, with no visible stop condition or bounded runtime.

Skill content
Loop forever → go back to step 4. Never stop polling.
Recommendation

Only enable continuous mode deliberately. Add a clear stop command, runtime limit, maximum order count, and user confirmation before ongoing background operation.

What this means

The agent may automatically deliver incorrect, unsafe, or unwanted work under the user's Atelier identity.

Why it was flagged

The skill chains external order input, local generation tools, upload, and delivery APIs without a visible human review or approval step.

Skill content
For each order: read `brief` → generate content with your available tools → `POST /upload` → `POST /orders/{id}/deliver`
Recommendation

Require user review before publishing services or delivering orders, and constrain what tools can be used based on the service category and customer brief.

What this means

Anyone who obtains the saved key may be able to control the user's Atelier agent, services, orders, or payout settings.

Why it was flagged

The skill stores a bearer API key persistently on disk. This is expected for the integration, but it is sensitive account authority.

Skill content
echo "ATELIER_API_KEY=$API_KEY" >> ~/.env
Recommendation

Store the API key in a secrets manager or restricted-permission file, avoid committing it to repositories, and rotate it if exposed.

What this means

Customer briefs and generated deliverables leave the local environment and are handled by Atelier's service.

Why it was flagged

The workflow exchanges briefs and deliverables through a third-party marketplace and may involve other agents; identity and data-boundary controls are not described in the visible artifact.

Skill content
sell creative services to humans and other agents ... upload to Atelier CDN, deliver
Recommendation

Do not include private or regulated data in briefs or deliverables unless Atelier's terms, privacy controls, and order identity are acceptable.

What this means

Users may need to install undeclared dependencies or run embedded code that was not separately packaged or scanned as a code file.

Why it was flagged

The embedded autonomous script imports a Python dependency, while the provided install information lists only curl and python3 and no install spec.

Skill content
import requests
Recommendation

Provide a reviewed install spec with pinned dependencies, or keep the skill to documented curl commands that require no extra packages.