learn-from

What problem does it solve?

This Skill solves the problem of turning diverse knowledge sources (papers, code repos, articles, or conversational notes) into vetted, actionable insights while enforcing credibility and security before any application or skill changes are made.

Core Features & Use Cases

• Source classification and routing: Automatically detect whether input is a paper, repository, article, or chat and invoke the correct sub-skill.
• Credibility and security gating: Run credibility checks and secure-* scans before extracting or applying findings.
• Insight extraction and application: Extract GOTCHAs, TECHNIQUEs, METRICs, and CONTRADICTIONs, recommend APPLY/PARTIAL/SKIP, and optionally create or update skills with validation.
• Contradiction resolution protocol: Present side-by-side comparisons and require explicit user decisions before replacing existing guardrails.
• Unified reporting: Produce structured reports with credibility score, security verdicts, extracted insights, and an impact summary for project application or skill library changes.
• Use case: Ingest an arXiv paper, run credibility scoring, extract reproducible techniques, and propose targeted skill updates with citations.

Quick Start

Ask the agent to "learn from" followed by a URL, file path, or pasted content to extract vetted insights and an application plan.