The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill performs file system operations restricted to the {project_root}/wenqu-skills/ directory. It reads draft content, versions, and changelogs, then writes cleaned output and metadata to a specific subfolder. This local file management is consistent with its stated purpose and does not access sensitive system paths or credentials.\n- [REMOTE_CODE_EXECUTION]: There are no attempts to download or execute external scripts, and no unverified dependencies are installed. Cross-skill calls to 'wenqu-image' are handled via the platform's standard tool invocation mechanisms.\n- [DYNAMIC_EXECUTION]: The skill uses template-based text cleaning and generation. While it can inline instructions from a related skill if the tool is unavailable, this remains within the vendor's local file structure and does not involve untrusted input or runtime compilation.\n- [CREDENTIALS_UNSAFE]: The documentation includes security-conscious guidance in 'references/auto-publish.md', explicitly warning against storing API keys or credentials in Markdown files to prevent accidental leakage in future integrations.

wenqu-publish