The Agent Skills Directory

[SAFE]: The build process relies exclusively on Node.js built-in modules (node:fs, node:path, node:url), avoiding third-party package dependencies and associated supply chain risks.
[SAFE]: The skill includes sanitization logic in its build script (build.mjs), using regular expressions to strip HTML tags from user-supplied text when generating Markdown documentation.
[SAFE]: External resource loading is limited to well-known services (Google Fonts) for visual styling of the generated HTML output.
[SAFE]: The skill instructions promote secure handling of sensitive documentation by recommending the use of git-ignored scratch directories for research and system artifacts.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data including vision docs and code surfaces which could contain malicious instructions.
Ingestion points: Vision documents and repository content (SKILL.md).
Boundary markers: Delegated tasks with shared briefs for subagents to provide primers (SKILL.md).
Capability inventory: Local Node.js execution and file system writes (build.mjs).
Sanitization: HTML tag stripping in build.mjs and entity escaping for packet payloads in template.html.

system-atlas