The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses spawnSync and spawn in scripts/figwright-operator.mjs to execute the MCP Inspector CLI via npx. It also uses execFileSync in scripts/webpage-to-figma.mjs to run npm for package installation and osascript to execute JavaScript for Automation (JXA) for macOS clipboard manipulation.
[EXTERNAL_DOWNLOADS]: The script scripts/webpage-to-figma.mjs dynamically installs the playwright-core package from the npm registry at runtime if it is not found in the local environment. Additionally, it provides an optional --official-script parameter that downloads a JavaScript file from a user-provided URL and saves it to a local cache directory (.cache/figma-official-capture/).
[REMOTE_CODE_EXECUTION]: The skill facilitates the execution of external code by downloading user-specified JavaScript and injecting it into a Chromium browser instance via Playwright to perform design capture tasks.
[DYNAMIC_EXECUTION]: The skill launches a browser instance using Playwright with the --disable-web-security flag enabled. This disables the Same-Origin Policy and other browser security boundaries, which is used to allow the capture runtime to access cross-origin assets and local files for design reproduction but increases the risk surface if malicious content is processed.
[OBFUSCATION]: The file assets/webpage-to-figma-runtime.js is a bundled and minified JavaScript file, making manual security auditing difficult. Deterministic detectors identified Base64 decoding patterns (atob) within this file, which appear to be used for legitimate processing of Data URLs and metadata during the clipboard capture process.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes content from external webpages provided via URL or HTML string. While focused on design extraction, this content enters the agent's context and could contain instructions targeting the LLM.
Ingestion points: scripts/webpage-to-figma.mjs (via --source parameter).
Boundary markers: None implemented for the ingested webpage content.
Capability inventory: File system access (writing screenshots/logs), network access (via Playwright), and system clipboard manipulation (osascript).
Sanitization: Content is processed by a browser runtime; no explicit sanitization of text content for prompt injection patterns was observed before it is returned to the agent context.

figma-content-operator