The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to launch a local Node.js dashboard server using the command node scripts/dashboard-server.mjs. This server provides localized file management and editing capabilities within the specified workspace.
[EXTERNAL_DOWNLOADS]: A version check feature is defined that uses curl or gh to query the author's repository (qin1473692580-ux/oh-story-claudecode) for release information.
[REMOTE_CODE_EXECUTION]: The update procedure utilizes npx skills add to download and install a ZIP archive from the author's official GitHub releases. This is a standard vendor-provided update mechanism for this ecosystem.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-generated novel text and project metadata which could potentially ingest malicious instructions.
Ingestion points: The skill reads novel content from designated project directories (e.g., 正文/, 大纲/) and uses it during routing decisions and exploration tasks.
Boundary markers: There are no explicit prompt delimiters or safety instructions used to wrap novel text when it is passed to routing logic or sub-agents.
Capability inventory: The skill can execute local Node.js scripts, perform shell-based network updates, and trigger sub-agents like story-explorer.
Sanitization: The dashboard server uses isPathInside with path normalization to prevent directory traversal, and the UI component escapes HTML characters when rendering Markdown previews.

story