idea-creator

Generate candidate research ideas from a literature review by identifying the review artifact to output a bounded set of 2 to 5 concepts with falsifiable claims and validation plans. This process isolates bootstrap concepts for later novelty filtering without triggering active experiments.

A bounded set of candidate ideas is a compact, verifiable collection of 2 to 5 research concepts generated from literature context. Each idea includes a stable identifier, title, falsifiable claim, literature-based novelty basis, validation plan, failure signals, and source references.

Yes, you need an existing literature review artifact before generating novel research ideas. The system requires this literature context to trace the novelty basis and source references for each candidate idea, ensuring an auditable trail for later novelty filtering.

Short-list research ideas for novelty filtering without starting an experiment by generating a bounded candidate set and writing outputs to the bootstrap workspace. This approach preserves traceability to the literature artifact while keeping bootstrap concepts isolated from active experiments.

Limitations include generating a maximum of 2 to 5 candidate ideas per run and restricting outputs to the bootstrap workspace. The process does not create an active experiment line, meaning generated concepts remain isolated until manually short-listed for novelty filtering.

Outputs created when turning a literature review into candidate ideas include an idea-candidates markdown file and an updated STATE.md file. These are written to the project bootstrap workspace to reflect the new candidate set and preserve traceability to the literature artifact.