The Agent Skills Directory

[SAFE]: The skill implements strict security protocols for credential management. It explicitly forbids the agent from reading, writing, or repeating sensitive secrets (like LARK_APP_SECRET and GMAIL_APP_PASSWORD) in the chat. It instructs the agent to check for the presence of secrets using non-revealing commands such as grep -q to avoid exposing values in the conversation context or shell history.
[SAFE]: The skill enforces a preflight check to ensure the .env file is included in .gitignore before any configuration occurs, preventing accidental leakage of credentials to version control systems.
[COMMAND_EXECUTION]: The skill orchestrates the execution of local repository scripts (e.g., node scripts/verify-setup.mjs and node scripts/setup-lark-base.mjs) to automate schema provisioning and verify connectivity. These scripts are part of the distributed repository and are used for their intended setup purposes.
[EXTERNAL_DOWNLOADS]: The skill informs the user that npm install will download standard dependencies, including a Chromium binary needed for PDF generation. These are standard operations for the described functionality.
[INDIRECT_PROMPT_INJECTION]: The skill defines a clear boundary for untrusted data by instructing the agent to extract specific tokens from user-provided URLs and use dedicated verification scripts rather than processing free-form input directly in a high-privilege context. Findings for this category are considered low risk due to the explicit delimitation of secret handling.

invoice-generator-guide