The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the tsx runner to execute local scripts for constructing the landing page and generating image placeholders. It requires file-write permissions to save the generated HTML and asset files to the filesystem.
[EXTERNAL_DOWNLOADS]: The scripts/imagegen.ts script fetches generated artwork from fal.ai's API. The generated HTML also includes a client-side component that retrieves repository statistics from the GitHub API. These interactions are limited to well-known technology services.
[DATA_EXFILTRATION]: Network operations are present for legitimate tool functionality, specifically for image generation and repository metadata retrieval. No unauthorized sensitive data access or exfiltration patterns were identified.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface through the brand brief input. User-supplied text is embedded directly into the generated HTML without sanitization, creating a potential XSS vulnerability.
Ingestion points: Data from the brand brief is stored in inputs.json and then processed by the composer script.
Boundary markers: Absent. The skill does not employ delimiters or security instructions to separate untrusted data from the HTML template.
Capability inventory: file-write is used to create the final index.html artifact.
Sanitization: Absent. The scripts/compose.ts renderer does not perform HTML entity escaping on user-supplied strings.

open-design-landing