Discover Research

Use this skill when the task is about finding, filtering, synthesizing, or operationalizing ideas for new <=2.1B LLM experiments.

Use it for

• Turning new papers, repos, and blog posts into candidate research ideas.
• Deciding which discovered ideas are worth turning into experiments.
• Bridging from research findings to architecture, data-mix, and evaluation decisions.
• Choosing which project skill to invoke next.

Local skill map

• paper-to-idea-extractor Use when new sources have been ingested and need to become structured candidate ideas.
• architecture-hypothesis-generator Use when a source-derived mechanism should become a concrete model/training hypothesis.
• dataset-mix-governor Use when a promising idea implies a data-mixture or synthetic-data decision.
• experiment-gatekeeper Use when an idea or prototype needs a go/no-go or promotion decision.
• experiment-report-writer Use when a run or evaluation pass needs a standardized summary.
• Use when the research loop should be packaged into repeatable CLI or daemon workflows.