The Agent Skills Directory

[SAFE]: The skill implements a content monitoring pipeline involving external data ingestion from well-known sources including YouTube, GitHub, Reddit, and standard RSS feeds. These operations are consistent with the skill's stated purpose of building intelligence pipelines.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes untrusted data from public web sources such as YouTube transcripts and Reddit posts.
Ingestion points: The discover.js, discover-rss.js, discover-github.js, and other adapters in scripts/content-monitor/ fetch content from external platforms.
Boundary markers: The pipeline architecture uses a structured JSON schema for the summarization phase, providing a data boundary between raw extraction and final evaluation.
Capability inventory: The skill utilizes yt-dlp for media transcription and makes network calls to established APIs (GitHub, Reddit, Google Search).
Sanitization: The multi-stage process (Extract -> Summarize -> Evaluate) reduces risk by transforming raw input into structured JSON summaries before final model reasoning.

content-monitor