The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process Markdown content from various sources to generate HTML reports, which constitutes an indirect prompt injection surface.
Ingestion points: Report source files (.md or .json) are read by the rendering script in scripts/render-core.mjs using fs.readFileSync.
Boundary markers: Not applicable as the skill functions as a file-to-file transformation utility rather than a prompt-interfacing agent.
Capability inventory: The skill uses standard Node.js file system capabilities (fs) to read sources and write HTML outputs.
Sanitization: The renderer implements robust sanitization including escapeHtml for all content, a whitelist-based scheme filter (normalizeTarget) that blocks dangerous protocols like javascript:, and a directory traversal check (isUnderRoot) that resolves symlinks via fs.realpathSync to ensure operations stay within the intended workspace.
[COMMAND_EXECUTION]: The skill instructions define specific CLI patterns for invoking its rendering engine via node. These commands are localized to the skill's own scripts and intended for processing document artifacts.
[SAFE]: The skill demonstrates defensive engineering through its inclusion of a dedicated security test suite (scripts/render-core.test.mjs) which specifically validates its protections against XSS, path traversal, and protocol-relative link attacks.

report-output-discipline