The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands to render content and manage deployments.
Runs node scripts/render-html.mjs <slug> or pnpm render <slug> to generate HTML artifacts.
Executes pnpm --filter recap-web build for web track builds.
Uses system-level openers (open, xdg-open, start) to display generated files to the user.
Executes pnpm deploy:preview for Vercel deployment upon user confirmation.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its automated research and content generation workflow.
Ingestion points: The research-scout subagent collects 5–12 external primary sources from the network during the topic research phase.
Boundary markers: The skill does not define explicit boundary markers or instructions to treat external source content as untrusted data, increasing the risk that instructions embedded in research sources could be obeyed by the agent.
Capability inventory: The skill has significant capabilities, including file-writing to the application source directory and shell execution via node and pnpm.
Sanitization: There is no evidence of sanitization or escaping of the data retrieved from external sources before it is passed to subagents responsible for code and content generation.
[DYNAMIC_EXECUTION]: The skill dynamically generates content and code that is then executed or rendered by local scripts.
Generates RecapPageContent JSON objects and SVG/Mermaid diagrams based on researched topics and audience levels.
Programmatically updates configuration files like active-content.json to route the application to the newly created content slug.
Utilizes a local script (scripts/render-html.mjs) to transform generated JSON into a standalone, inlined HTML file at runtime.

recap-topic