The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses sensitive files in the user's home directory to load configuration and secrets.
Evidence: In __init__.py, the load_config function attempts to read from ~/.openclaw/.env and ~/.claude/feishu-config.env. These paths are identified as sensitive file locations.
[CREDENTIALS_UNSAFE]: The skill extracts and uses application secrets (App ID and App Secret) from local environment files.
Evidence: The load_config function specifically parses FEISHU_APP_ID and FEISHU_APP_SECRET from .env files. While necessary for the skill's primary function of interacting with the Feishu API, accessing generic environment files increases the risk of credential exposure.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it fetches and processes content from external URLs without sanitization or boundary markers.
Ingestion points: The doc_to_md and url_to_md functions in __init__.py ingest content from Feishu documents and external URLs (e.g., WeChat).
Boundary markers: No delimiters or instructions to ignore embedded commands are present in the prompts.
Capability inventory: The skill utilizes the requests library to make network calls and fetch content, but does not currently implement dangerous execution capabilities like subprocess or eval.
Sanitization: The skill performs basic Markdown formatting but does not sanitize the input text for potential malicious instructions targeted at the consuming agent.

feishu-doc-converter