The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill makes network requests to cubox.pro to fetch user articles and metadata.- [EXTERNAL_DOWNLOADS]: The script scripts/cubox.ts identifies image URLs within article content and downloads them to the system's temporary directory (os.tmpdir()). These URLs are controlled by external content and are not restricted to trusted domains.- [DATA_EXFILTRATION]: The skill reads the CUBOX_AUTHORIZATION credential from a local .env file and transmits it to the Cubox API in request headers.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to the processing of untrusted web content.
Ingestion points: Article HTML fragments and metadata are retrieved from cubox.pro via the scripts/cubox.ts script.
Boundary markers: The generated Markdown files containing article content do not include delimiters or instructions for the agent to ignore embedded commands.
Capability inventory: The agent is explicitly instructed in SKILL.md to "Analyze the exported Markdown" and synthesize findings.
Sanitization: The skill performs basic HTML sanitization by removing script and style tags and converting content to Markdown using the turndown library.- [COMMAND_EXECUTION]: The skill uses bun run to execute local TypeScript scripts for API interaction and content processing.

cubox-research