The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using npx to initialize the environment and add the MCP server. These commands are intended to be executed by the agent if shell access is available or by the user manually.
[EXTERNAL_DOWNLOADS]: The setup process requires downloading the @feimacode/open-design-agent-kit package from the npm registry and references binary releases from the vendor's GitHub repository.
[REMOTE_CODE_EXECUTION]: The instructions to run npx commands for tool initialization and server registration involve downloading and executing remote code from the npm registry.
[INDIRECT_PROMPT_INJECTION]: The skill features a significant attack surface for indirect prompt injection by processing untrusted data from external sources.
Ingestion points: The agent is instructed to read and extract content from various document formats (PDF, DOCX, XLSX, CSV, Markdown) passed as sources. It also ingests user feedback and comments retrieved from a browser preview.
Boundary markers: The instructions do not define specific delimiters or warnings to isolate ingested document content from the agent's core instructions.
Capability inventory: The agent is authorized to write files, execute shell commands, and interact with external design platforms via tools.
Sanitization: There are no documented steps for sanitizing or validating extracted document content before it is used for design generation.
[DATA_EXFILTRATION]: The publish_open_design_artifact tool allows the agent to host artifacts on external providers. This mechanism could be used to exfiltrate sensitive information extracted from workspace documents if the agent incorporates that data into a published artifact.
[DYNAMIC_EXECUTION]: The skill authors and renders self-contained HTML/JavaScript code (including WebGL shaders and complex layouts) to perform visual checks, generate screenshots, and provide live previews.

open-design