The Agent Skills Directory

[SAFE]: The skill implements a robust business logic workflow that requires verification of procurement approval status (e.g., checking for 'APPROVED' status and date ranges) before proceeding with reimbursement, reducing the risk of unauthorized financial requests.- [SAFE]: Sensitive data handling is explicitly addressed; the instructions mandate checking for and redacting card secrets, tokens, or irrelevant private information from receipt images or PDFs prior to any network transmission.- [SAFE]: Configuration management follows industry standards, utilizing local application-specific directories ($XDG_CONFIG_HOME/lark-reimbursement/ or ~/.config/lark-reimbursement/) for storing tenant-specific URLs and private task states.- [SAFE]: Interaction with external services is performed through structured API calls using the lark-cli tool. The documentation specifically advises using JSON serialization and argument arrays to prevent command injection when handling dynamic file names.- [SAFE]: The skill utilizes a local private directory (./private-task) for temporary file storage and task recovery, ensuring that sensitive transaction data and API receipts are not stored in shared or public locations.

lark-reimbursement