The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill performs a version check by fetching metadata from the vendor's official GitHub repository (treehausdev/skills). It also provides instructions for installing the skill via 'npx skills add treehausdev/skills', which fetches resources from the vendor's infrastructure.
[COMMAND_EXECUTION]: The skill automates the development lifecycle by executing various CLI commands. This includes scaffolding projects with 'create-next-app', managing Git repositories and GitHub authentication with 'gh', and handling deployment and domain configuration via the 'vercel' CLI.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it crawls and extracts data from external websites provided by the user during the intake phase (Phase 1, Step 3).
Ingestion points: Website crawling of sitemaps and page content (Home, About, Contact, etc.) to extract business information, contact details, and brand assets.
Boundary markers: No explicit delimiters or instructions to ignore embedded instructions are defined for the data extracted from the crawled pages.
Capability inventory: The agent possesses significant capabilities, including file system access, network requests, and the ability to execute shell commands ('npx', 'git', 'gh', 'vercel').
Sanitization: There is no evidence of sanitization or validation of the content crawled from external sites before it is interpreted by the agent for project generation.

treehaus