The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The file scripts/feishu_doc_reader.py contains hardcoded sensitive app_secret values for two Feishu tenants: "hanxing" (5djHWjk8t6QHRquDJXG9JiNEgPynmnIN) and "personal" (bX21pNOyAXHwFXWf0SVGphWQrqrgC5Gd). These secrets are used to programmatically obtain tenant_access_token values, granting the skill full access to the linked Feishu applications.
[COMMAND_EXECUTION]: The script scripts/feishu_doc_editor.py implements a destructive automation routine designed to clear Feishu documents. The clear_document function uses Playwright to simulate a loop of 50 consecutive "Control+A" and "Backspace" keystrokes. This aggressive behavior poses a high risk of accidental or malicious data loss.
[PROMPT_INJECTION]: The skill exhibits a significant attack surface for indirect prompt injection by processing untrusted external content and utilizing it to perform document modifications.
Ingestion points: Raw document content is fetched from remote Feishu document tokens using the API in scripts/feishu_doc_reader.py.
Boundary markers: The skill lacks any delimiters or safety instructions that would prevent the agent from obeying instructions embedded within the document content being "optimized".
Capability inventory: The skill has highly permissive capabilities, including the ability to completely delete document content and replace it with new text via the scripts/feishu_doc_editor.py browser automation script.
Sanitization: There is no evidence of input validation, instruction filtering, or sanitization of the fetched document content before it is processed by the language model.

feishu-doc-optimizer