Generate and select literature-grounded research directions from an active baseline and codebase.
npx skills add https://github.com/OpenLAIR/dr-claw --skill ds-idea
Please help me install this Agent Skill. Skill: ds-idea Source: https://github.com/OpenLAIR/dr-claw/tree/main/skills/ds-idea Command: npx skills add https://github.com/OpenLAIR/dr-claw --skill ds-idea
Research ideation often collapses onto the first plausible idea or produces vague inspiration that cannot survive novelty, feasibility, and manuscript-defensibility checks. This Skill turns an active baseline, codebase, and evaluation contract into concrete, testable, literature-grounded research directions with explicit tradeoffs.
Use the ds-idea skill to analyze the current baseline's limitations, survey related work, and select the next research direction for this quest.
High-intent search queries and answers about installing and using this skill.
Start from a concrete limitation of the baseline, split symptom from mechanism hypothesis, then run a bounded divergent pass producing 6-12 raw ideas across 2-4 ideation lenses. Converge to 2-3 serious candidates and select one with an explicit evidence-based decision path.
The survey must durably cover at least 5 and usually 5-10 task-relevant or mechanism-relevant papers before any direction is promoted. If the direct neighborhood is smaller, fill remaining slots with the closest adjacent translatable work and record that explicitly.
Skip it when the baseline gate is unresolved, the problem framing is still unclear, or the next step is obviously write-up or finalization. In those cases route back to baseline or scout stages first.
Build a closest-prior-work table comparing mechanism, task, dataset, and metric overlap, then assign each candidate a verdict of novel, incremental but valuable, or not sufficiently differentiated. Never claim novelty without a written related-work comparison.
It produces a limitations analysis, literature survey report, related-work map, candidate list, selected idea draft, and optionally a paper outline seed. The handoff includes a falsifiable claim, metric key, minimal experiment, abandonment condition, and strongest alternative hypothesis.
ResearAIStandardize a reference object for quest evaluation with plan and checklist requirements.
Rycen7822Plan and execute auditable experiments with PLAN.md and CHECKLIST.md.
RedWoodOGQuery the arXiv REST API to retrieve paper metadata and abstracts.
weathonChecks whether a research idea exists in alphaXiv literature via structured reporting and caching.
