The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/enrich_via_wechat_web_api.py utilizes the Chrome DevTools Protocol (CDP) to execute dynamically constructed JavaScript within an active browser session. This involves using Runtime.evaluate to inject code into the WeChat editor to manipulate fields not available via the public API.
[DYNAMIC_EXECUTION]: The script scripts/publish_via_gateway.py dynamically loads the env_manager.py module from variable filesystem paths (e.g., ~/.agents/skills/lov-env-management/scripts/env_manager.py). Loading executable code from computed paths in the user's home directory is a high-risk pattern.
[COMMAND_EXECUTION]: The script scripts/publish_existing_draft.py uses subprocess.run to call the macOS security utility. This allows the skill to retrieve stored credentials directly from the system keychain.
[DATA_EXFILTRATION]: The skill transmits article content and sensitive service credentials (AppSecret) to the vendor's gateway at api.lovstudio.ai. While documented as a feature to bypass IP whitelisting, it involves sending secrets to a third-party server.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted Markdown and HTML files as ingestion points. It lacks explicit boundary markers or sanitization to prevent the agent from potentially following instructions embedded within the article content, which could lead to unauthorized actions given the skill's broad capabilities.
[CREDENTIALS_UNSAFE]: The skill manages authentication by retrieving secrets from multiple sources, including local environment variables and the system keychain, which increases the surface area for credential exposure.

lov-publish-wechat-article