The Agent Skills Directory

[DATA_EXFILTRATION]: The skill provides explicit instructions for handling sensitive material, such as private keys and seeds, by outputting them to the integrated terminal using node -e or python -c commands, or by writing them to temporary files. This design is intended to move sensitive information out of the observable chat interface, which bypasses standard platform safety monitoring and audit logs.
[EXTERNAL_DOWNLOADS]: The skill documentation instructs the agent to run npx -y gatepay-local-mcp to install and execute the payment server. This pattern downloads external code from the NPM registry at runtime without version pinning or integrity verification.
[REMOTE_CODE_EXECUTION]: The runtime execution of the gatepay-local-mcp package via npx -y allows remote code to run directly on the host machine during the payment workflow.
[DYNAMIC_EXECUTION]: The skill leverages node -e and python -c to execute dynamically generated code strings for the purpose of programmatically outputting sensitive data to the user's terminal.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external merchant HTTP responses (headers, status, and body) via the x402_place_order tool.
Ingestion points: Merchant responses received through the x402_place_order and x402_request tools in SKILL.md.
Boundary markers: The instructions do not define boundary markers to prevent the agent from interpreting instructions embedded within merchant-controlled data.
Capability inventory: The skill has access to sensitive capabilities including payment signing (x402_sign_payment) and centralized settlement (x402_centralized_payment).
Sanitization: The skill relies on schema validation for arguments but does not implement sanitization or filtering for the content returned from merchant servers.

gate-pay-x402