The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple external sources and local configuration files, which could be leveraged to influence the agent's behavior.
Ingestion points: Remote component registry (https://zardui.com/r), project-specific documentation (https://zardui.com/docs/components/*.md), and the local components.json configuration file.
Boundary markers: The instructions do not specify the use of clear delimiters or guardrails when the agent is interpreting documentation or registry metadata to generate code.
Capability inventory: The skill has access to shell execution via Bash (running npx zard-cli) and can write component source code directly into the user's project directory. The MCP server provides tools like install-component which can modify the project environment.
Sanitization: The skill assumes the integrity of the data provided by the registry and the project's own configuration without explicit validation steps described in the instructions.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of code and configuration metadata from a remote repository.
Evidence: The CLI and MCP tools fetch component source code and JSON metadata from https://zardui.com/r.
Context: This behavior is central to the skill's purpose as a component library manager, using the vendor's official registry to provide project updates.
[COMMAND_EXECUTION]: The skill utilizes shell commands to perform project initialization and component management tasks.
Evidence: The allowed-tools section grants access to Bash for execution of npx zard-cli, pnpm dlx zard-cli, and bunx zard-cli commands.
Context: These commands are used to automate project setup, add components, and provide diagnostic information about the local environment.

zard