The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the shell command defined in the dev.command field of the .chalk/chalk.json configuration file to start the application's development server.
[EXTERNAL_DOWNLOADS]: Fetches browser binaries from Microsoft's infrastructure using the npx playwright install chromium command.
[REMOTE_CODE_EXECUTION]: Uses npx to dynamically download and run the tsx and playwright Node.js packages at runtime.
[REMOTE_CODE_EXECUTION]: Writes a temporary TypeScript script to .chalk/local/capture-script.ts and executes it to perform browser automation tasks.
[PROMPT_INJECTION]: Exposed to indirect prompt injection through project-level data sources.
Ingestion points: Processes configuration from .chalk/chalk.json, package.json, and application source code via grep and glob patterns.
Boundary markers: No explicit delimiters or instructions are used to separate untrusted project content from agent logic.
Capability inventory: Has permissions for Bash execution, file Write, and network access via curl and npx.
Sanitization: No validation or escaping is performed on data retrieved from project files before it is used in shell commands or script templates.

capture-pr-visuals