The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python's subprocess and Node.js's spawn to execute shell commands, including running its internal CLI, project tests, and system-level path openers. Evidence found in scripts/cli2anything.py and bin/cli2anything.mjs.
[DYNAMIC_EXECUTION]: The tool programmatically generates task-focused Node.js packages and uses npm link to globally register them on the host system. Evidence in bin/cli2anything.mjs.
[DYNAMIC_EXECUTION]: The skill implements a 'Browser Session Mode' using the Chrome DevTools Protocol (Runtime.evaluate) and a custom browser extension bridge to execute logic directly within active browser tabs. While intended for API discovery, this represents a high-privilege execution capability. Evidence in src/browser-cdp.mjs and bin/cli2anything.mjs.
[EXTERNAL_DOWNLOADS]: The skill fetches JavaScript and JSON data from external domains (e.g., cdn.marmot-cloud.com, api.dshfind.com) to analyze and probe API endpoints. Evidence in bin/cli2anything.mjs and src/dshfind-target.mjs.
[INDIRECT_PROMPT_INJECTION]: The skill provides an interactive 'Drilldown' interface and an 'AI Sidebar' that ingest and render data directly from external API responses. This creates an attack surface where untrusted data could influence agent actions or exploit local capabilities.
Ingestion points: src/drilldown-html.mjs (renders API responses) and src/zenmux-browser-client.mjs (fetches data).
Boundary markers: The skill applies HTML entity escaping (escapeHtml) to data rendered in the UI.
Capability inventory: Extensive local execution rights via subprocesses, file writes, and browser script injection.
Sanitization: Uses a dedicated escapeHtml utility for UI data interpolation.

lov-cli2anything