The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a helper script, scripts/render-reader-formats.ts, which is executed by the agent to generate document previews. This script spawns local browser processes (Chrome/Chromium) to print Markdown files to PDF and uses system-default openers (e.g., open, xdg-open, cmd) to display the results to the user.
[EXTERNAL_DOWNLOADS]: The rendering utility fetches the Mermaid diagramming library from the jsDelivr CDN (cdn.jsdelivr.net) to render flowcharts and timelines within generated HTML and PDF files.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data provided by the user to populate business case configuration and metadata.
Ingestion points: User-provided facts and messages, as well as the internal/intake.yaml file, are used to scaffold the case project.
Boundary markers: The instructions rely on a strict state-machine workflow (asking one fact at a time) to manage input, though they do not specify explicit text delimiters.
Capability inventory: The skill can write files to the case-<slug>/ directory and execute the provided rendering script which interacts with the local filesystem and browser.
Sanitization: The rendering script includes sanitization logic, such as escapeHtml for content rendering and a safeUrl check that restricts link protocols to http, https, and mailto to prevent script injection via Markdown links.

case-bootstrap