The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill utilizes tools like inspect_document and get_story to read the contents of workspace documents. This creates a surface for indirect prompt injection if those documents contain malicious instructions designed to manipulate the agent.
Ingestion points: document content read via inspect_document and get_story (SKILL.md).
Boundary markers: None present; instructions do not specify the use of delimiters or warnings to ignore instructions within the processed data.
Capability inventory: The agent can execute shell commands (node), write files (save_document), and open UI panels (open_editor).
Sanitization: No explicit sanitization or filtering of document content is mentioned in the instructions.
[COMMAND_EXECUTION]: The skill instructs the agent to run a local Node.js script (node ui/dist/adapter.mjs --serve) to manually start the editor backend service if the tool is not automatically registered.
[DYNAMIC_EXECUTION]: The skill refers to a build process for ui/dist/adapter.mjs and instructs the agent to execute this generated file to facilitate the editor's UI and automation capabilities.

designcraft-editor