Paper To Idea Extractor

When to use this skill

• You ingested new papers/posts/repos and need candidate ideas for the idea box.
• You need structured idea atoms instead of unstructured notes.
• You need provenance-linked hypotheses for automated architecture generation.

Inputs required

• Source content summary or full text chunks.
• Source metadata (url, date, type, title).
• Current constraints (param budget, runtime constraints, tokenizer).

Workflow

1. Parse source into idea atoms:
   • mechanism
   • expected benefit
   • compute cost/risk
   • compatibility constraints
2. Score each atom for novelty and feasibility under current constraints.
3. Remove weak or non-testable atoms.
4. Consolidate near-duplicates and preserve alternate-source provenance.
5. Emit idea-box-ready entries with deterministic fields.