The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to use npx @superdoc-dev/cli@latest to run the SuperDoc tool. This retrieves and executes code from the NPM registry at runtime. The package is a scoped vendor resource associated with the author's platform.
[COMMAND_EXECUTION]: The instructions include bash scripts to check for and execute local text editors (code, code-insiders, cursor) to provide a live preview of document changes if running within a VS Code environment. It also executes various subcommands of the superdoc CLI.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes external content from user-provided .docx files through commands like get-text and get-markdown.
Ingestion points: Document reading operations defined in references/reading.md (get-text, get-markdown, get-html, find, query match).
Boundary markers: None identified. The instructions do not specify the use of delimiters or 'ignore' instructions for the content extracted from Word files.
Capability inventory: The skill has significant capabilities, including file system write access (save), shell command execution (superdoc, npx, editor binaries), and network access features for collaboration.
Sanitization: No sanitization or filtering logic is present in the skill instructions to handle potentially malicious instructions embedded in the documents being processed.

Word Editor