The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local subprocesses to manage searches and fallback logic.
The search_news.js script employs child_process.execFile to invoke bash or node for fallback search operations via a sibling web-search skill.
The search-news.sh and search-news.ps1 entry points facilitate execution of the core Node.js logic.
[EXTERNAL_DOWNLOADS]: The skill fetches data from numerous external technology platforms and community forums.
It communicates with reputable news sources including TechCrunch, The Verge, Wired, and Hacker News via RSS and API endpoints.
The network_detector.js script performs connectivity checks against well-known services like Cloudflare to determine the available network environment.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of external news summaries.
Ingestion points: External article titles and summaries are ingested through scripts/parsers/rss_parser.js and scripts/parsers/hn_parser.js.
Boundary markers: The output formatting instructions in SKILL.md do not implement explicit delimiters or safety warnings for the agent when interpreting external content.
Capability inventory: The skill possesses capabilities for subprocess execution via execFile, network requests via the https module, and file system writes for caching and temporary storage.
Sanitization: While rss_parser.js includes basic HTML tag removal, it does not filter for semantic instructions or malicious prompts embedded within news summaries.

technology-search