The Agent Skills Directory

[SAFE]: The skill instructions and templates do not contain malicious patterns, obfuscation, or unauthorized data exfiltration logic. It follows the principle of least privilege by restricting access to sensitive credentials.
[COMMAND_EXECUTION]: The skill utilizes the airwallex CLI and uuidgen utility. These are legitimate tools for interacting with the vendor's platform and generating unique identifiers for financial transactions.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user-provided documents like spreadsheets or lists to provision cards.
Ingestion points: SKILL.md (Phase 1, Step 1) describes extracting specifications from external documents.
Boundary markers: The skill lacks formal delimiters but utilizes structured data tables for user verification.
Capability inventory: The agent can create cardholders, issue cards, and modify spend limits using the Airwallex API.
Sanitization: Strong procedural controls are in place; the agent is required to present extracted data to the user and obtain explicit confirmation for every individual JSON payload before execution.

card-provisioning