Research Ideation: From Scope to Ranked Proposals

Where this fits

You're a quant researcher staring at a blank page. You have a vague sense that you want to contribute something novel, but you don't know where to start. The research-ideation skill turns that open-ended goal into a structured pipeline: it helps you define your scope, anchor your thinking to existing literature, generate a single-core idea, iterate, and finally rank competing ideas with an ELO tournament—all before you commit to writing a full proposal.

Why agents benefit

• Anchor-first grounding forces the agent to read and cite a specific paper before generating ideas, preventing hallucinated novelty and keeping contributions incremental.
• Three-stage scope selection narrows the problem space (e.g., from broad field to specific subproblem) so the agent doesn't waste tokens exploring dead ends.
• ELO tournament ranking uses a transparent formula (Final = N+R+C−D) to compare ideas on novelty, relevance, clarity, and difficulty—giving you a defensible decision metric.
• Evo-memory updates persist the chosen direction and lessons learned, so future ideation sessions get smarter without re‑reading the same papers.
• Manuscript-quality expansion turns the winning idea into a structured proposal, ready for a paper planner or drafting tool.

Practical scenarios

Finding a new research direction from scratch

You have a domain (e.g., volatility modeling) but no specific gap. The skill runs scope selection, pulls candidate anchors from your local paper library, and generates three to five distinct ideas. The tournament ranks them, and you pick one to expand.

Evaluating novelty of a half-formed idea

You already have a hunch—say, applying a transformer to a classic pricing model. The skill grounds your idea against a chosen anchor paper, scores it against the formula, and flags whether it's too incremental or too disconnected from existing work.

Comparing multiple ideas before a deadline

You need to choose between three possible research threads for a conference submission. The skill runs ELO matches on pairwise comparisons, producing a ranked list with reasoning you can present to collaborators.

Add it to your agent workflow

Install the skill with:

npx skills add CamusGIT/EvoQuant --skill research-ideation

Then invoke it in an agent prompt like:

Use research-ideation to find a novel approach for improving option pricing under jump diffusion. Start with scope selection, then anchor to a recent paper from my local library.

The skill will produce a ranked set of ideas and a proposal template. You can also combine it with other skills from the same repo—for example, local-paper-navigator to fetch papers, or research-survey if you need a full literature review first.

Compare before adopting

Before committing, compare research-ideation against alternatives in three dimensions:

• Quality signals: Does the output actually cite real papers? Are the scoring criteria transparent and reproducible?
• Maintenance freshness: Check the GitHub repo's commit history and issue tracker—does the author actively update it for new LLM APIs or agent frameworks?
• Workflow fit: The skill is deliberately scoped to ideation, not paper search, survey, or planning. If your pipeline already covers those steps, verify this skill doesn't duplicate or conflict with them.

Why it is worth tracking

research-ideation stands out because it enforces a rigorous, repeatable process—not just a fancy prompt. The ELO tournament and evo-memory are concrete mechanisms that improve with use, and the project has 212 stars with a clear v3.0.0 release. It's worth tracking if you build agents for research-heavy domains beyond quant: the pipeline generalizes to any field where you need anchor-grounded, ranked idea generation. For a broader view of this skill and its ecosystem, see https://www.openagentskill.com/skills/camusgit-research-ideation. Evaluate it when your agent needs to move from “brainstorm” to “decision” without losing rigor.