The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/export-to-html.py interacts with the wpsnote-cli binary using the subprocess.run function.
Evidence: Implementation found in the _cli_run function.
Context: Arguments are passed as a list and the shell is disabled, which mitigates shell injection risks. This is standard functionality for retrieving note data from the vendor's application.\n- [PROMPT_INJECTION]: The skill processes external data from WPS notes, which creates a potential surface for indirect prompt injection.
Ingestion points: Note content retrieved via the wpsnote-cli tool or MCP tools as defined in scripts/export-to-html.py.
Boundary markers: No delimiters or specific instruction-ignore warnings are applied to the note content before processing.
Capability inventory: The skill can execute wpsnote-cli and write to the local file system to generate HTML files.
Sanitization: The conversion logic uses html.escape to ensure HTML safety, but does not sanitize content against LLM-specific instruction overrides.\n- [SAFE]: The skill follows secure coding practices for its internal operations.
Evidence: Configuration files are parsed using yaml.safe_load() in both scripts/export-to-html.py and scripts/md_to_html.py, preventing unsafe deserialization.
Evidence: Network activity is limited to referencing public image URLs in the generated HTML, and local file access is restricted to reading the note's XML or the user's provided Markdown files.

wechat-publisher