The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions and reference materials include multiple references to the domain 'project.dev', which has been flagged as malicious by automated scanners. These URLs are included in landing page templates and documentation site configuration as example canonical links and social metadata targets. Evidence: 'https://project.dev/', 'https://project.dev/social-card.png', and 'https://project.dev/docs/getting-started/installation/'.\n- [REMOTE_CODE_EXECUTION]: The skill provides build pipeline configurations (static-site-tooling.md) that fetch and execute external code from the official GitHub Actions organization to facilitate deployment. This includes actions for checking out code, configuring pages, and deploying artifacts to GitHub Pages.\n- [COMMAND_EXECUTION]: The skill contains instructions for the agent to recommend and configure various shell-based build commands for static site generators. Examples include 'hugo --minify', 'npx astro build', and 'npx @11ty/eleventy'. These commands facilitate the transformation of source content into a deployable site.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests user-provided site descriptions and content to generate complex code and configuration. \n
Ingestion points: User requests regarding site architecture and content types (SKILL.md). \n
Boundary markers: No explicit delimiters or safety instructions are provided to separate user-supplied text from the generation logic. \n
Capability inventory: The agent can generate interactive scripts (JavaScript for clipboard and theming) and automated build workflows (GitHub Actions). \n
Sanitization: No specific filtering or validation is described to ensure user content does not inject malicious components into the resulting website.

using-site-designer