direction-narrowing
$
npx mdskill add yogsoth-ai/de-anthropocentric-research-engine/direction-narrowingFocus within chosen field(s). Identify specific sub-directions and present ranked candidates.
SKILL.md
.github/skills/direction-narrowingView on GitHub ↗
--- name: direction-narrowing description: Focus within the user's chosen field(s). Identify specific sub-directions through deep paper and web research, then present ranked candidates. Use after landscape-reconnaissance has identified fields of interest. --- # Direction Narrowing Focus within chosen field(s). Identify specific sub-directions and present ranked candidates. ## Available SOPs | SOP | Purpose | Execution | |-----|---------|-----------| | broad-paper-search | Scan papers in the chosen field(s) | import: literature-overview | | deep-web-search | Deep reading of web resources in the field | subagent | | present-candidates | Present ranked sub-directions to user | dialogue | ## Methodology Guidance - hot-start may only need partial SOP execution (a few searches for context) - You decide search depth based on information sufficiency - `present-candidates` depth scales by start mode: - cold-start: broad sub-directions available to pursue - warm-start: specific sub-problems and research tracks - hot-start: granular knowledge points, technical details ## Hard Constraints - `broad-paper-search`: at least 80 papers scanned - `deep-web-search`: at least 30 web pages read in full ## Output (Tactic-Level Aggregation) `RankedCandidates[] + user's selection`