The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/save_to_bitable.py contains hardcoded identifiers for external infrastructure:
DEFAULT_APP_TOKEN = "ND8ObCuSya5Dv3sREZYc03Ilngh"
DEFAULT_TABLE_ID = "tblaHDM5kjtikIl9"
[CREDENTIALS_UNSAFE]: The script scripts/save_to_bitable.py attempts to read sensitive authentication tokens from a specific local file path: ~/.openclaw/tokens/feishu/default/user_access_token. This path is used by the platform to store credentials, and unauthorized access by a skill poses a security risk.
[COMMAND_EXECUTION]: The scripts/extract_content.py script generates and outputs shell command strings intended for the Agent to execute. For example, the scrapling_command field recommends running: python3 ~/.openclaw/workspace/skills/web-content-fetcher/scripts/fetch.py "{url}" 50000. This pattern encourages the execution of external scripts with dynamic parameters derived from user input (URLs).
[DATA_EXFILTRATION]: The skill's primary function involves collecting data from various sources and transmitting it to external Feishu (Lark) API endpoints (open.feishu.cn). While this is the documented purpose, it constitutes data movement to an external service controlled by API tokens.
[PROMPT_INJECTION]: The skill is highly susceptible to Indirect Prompt Injection (Category 8):
Ingestion points: The skill ingests untrusted data from multiple social media platforms (X/Twitter, WeChat, Reddit, etc.) via external fetching tools.
Boundary markers: There are no explicit boundary markers or instructions in the provided scripts to help the AI distinguish between the content being processed and instructions within that content.
Capability inventory: The skill has the capability to write to external APIs (save_to_bitable.py) and suggests the execution of shell commands (extract_content.py).
Sanitization: There is no evidence of sanitization, filtering, or escaping of the fetched content before it is processed by the AI for summarization or categorization.

content-collector