Research

Create and manage experimental features in the Doodle project. Use this skill when user requests creating a new doodle, adding features to the features folder, or mentions working on a specific language-based experimental project. Handles both new feature creation and continuation of existing features.

Automatically applies when evaluating LLM performance. Ensures proper eval datasets, metrics computation, A/B testing, LLM-as-judge patterns, and experiment tracking.

技術サイトから特定の技術に関する記事・動画・スライド・論文を抽出・分析します。Qiita、Zenn、Stack Overflow、GitHub、Dev.to、Medium、YouTube、arXiv等50以上のソースから最新記事を検索・フィルタリングし、トレンド分析やキャッチアップに活用できます。

Access and use project memories from previous sessions. Use this skill when you need to recall past decisions, check project conventions, or understand user preferences that were established in earlier sessions.

Analyze competitive courses, industry trends, learner demand, pricing, and technology landscape for educational market positioning. Use when launching new courses, competitive analysis, or market validation. Activates on "market research", "competitive analysis", "course demand", or "pricing research".

Edit writing to adopt Paul Graham's exceptionally clear style for research and long-form content - concrete language, varied sentence rhythm, accessible formality, specific evidence. Use when user wants to rewrite content in PG's style or asks to "make this clearer" or "simplify research writing."

Analyzes errors, searches past solutions in memory, provides immediate fixes with code examples, and saves solutions for future reference. Use when user says "debug this", "fix this error", "why is this failing", or when error messages appear like TypeError, ECONNREFUSED, CORS, 404, 500, etc.

Citation formatting rules for law review articles using The Bluebook (21st ed.) academic style

Convert experiment results to Markdown research notes and optionally update research state with evidence

Enhance book notes with Blinkist-style summaries. Use when asked to "enhance notes", "improve book notes", "add key insights", "expand notes", or "make notes better". Adds Core Message, Key Insights, Notable Quotes, and Who Should Read sections via parallel web research.

Tum helper fonksiyonlari referansi. Use for quick lookup of available functions.

Loop 2 of the Three-Loop Integrated Development System. META-SKILL that dynamically compiles Loop 1 plans into agent+skill execution graphs. Queen Coordinator selects optimal agents from 86-agent registry and assigns skills (when available) or custom instructions. 9-step swarm with theater detection and reality validation. Receives plans from research-driven-planning, feeds to cicd-intelligent-recovery. Use for adaptive, theater-free implementation.

Propose next mechanistic interpretability experiments based on research state, hypotheses, and existing evidence

Phase 2 of disciplined development. Creates implementation plans based onapproved research. Specifies file changes, function signatures, test strategy,and step sequence. Requires human approval before implementation.

Troubleshoot and manage GCP e2-micro VM running eth-realtime-collector. Use when VM service is down, systemd failures occur, real-time data stream stops, or VM network issues arise. Keywords systemd, journalctl, eth-collector, gcloud compute.

Create, improve, and debug professional LuaLaTeX documents using KOMA-Script + the moderndoc package (and the legacy modern-doc shim). Use for: generating new documents from a goal, improving typography/layout, producing PDF/A-4-friendly output, setting up minted/biblatex/tabularray/tikz, converting Mermaid → TikZ, and debugging compilation logs.

Specialized skill for editing and refining EAD2002 finding aids for the Terrence Kaufman Papers at the Archive of Indigenous Languages of Latin America (AILLA), Lillas Benson Latin American Studies and Collections, University of Texas at Austin. Use when working with KaufmanEAD.xml, implementing staff feedback, validating against EAD2002 standards, or referencing institutional finding aid examples.

Conduct parallel multi-angle research on programming topics. Break down complex research questions into 3-10 specific investigation avenues, delegate each to parallel sub-agents, and synthesize findings with working code examples and sources. Use when investigating error messages, comparing solutions, or exploring implementation approaches.

Use this skill when the user mentions "plan", "track", "spec", "conductor", asks about project progress, or references conductor files like plan.md, spec.md, tracks.md, product.md, tech-stack.md, or workflow.md. Also use when the user wants to understand the Conductor methodology or Context-Driven Development workflow.

Enforces visibility pattern for testability by detecting private methods that prevent unit testing (causes 20-40% coverage loss). References authoritative style guides rather than duplicating. Auto-refactors code to private[module] object pattern with user approval. Detects inline lambdas and suggests extraction to named functions ("laws"). Blocks during scala-ninja review (Phase 2) as peer review rejection point.