The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (MEDIUM): The script scripts/create-onchain-app.py executes npm create onchain@latest. This command downloads and runs a project generator from the npm registry, which constitutes remote code execution. Since the provider (Coinbase) is not on the predefined trusted list, this is flagged as a risk.
[COMMAND_EXECUTION] (MEDIUM): The scripts/validate-setup.py script invokes npm run build. This command executes arbitrary scripts defined in the project's package.json. If the project files are untrusted or modified, this can lead to arbitrary code execution on the host system.
[PROMPT_INJECTION] (LOW): The skill accepts user-provided project names which are then interpolated into shell commands. While alphanumeric sanitization is performed, this represents an indirect injection surface.
Ingestion points: scripts/create-onchain-app.py via the project_name argument.
Boundary markers: Absent.
Capability inventory: Shell command execution via subprocess.run(shell=True) and project structure modification.
Sanitization: Input is validated to be alphanumeric with hyphens and underscores allowed.
[EXTERNAL_DOWNLOADS] (LOW): The script scripts/setup-environment.py performs npm install @coinbase/onchainkit, which downloads and installs external packages from the npm registry during the environment setup process.

onchainkit