研究

Systematic data validation, error detection, cross-source reconciliation, and query correctness checking for analytical work. Use when validating Snowflake queries, catching calculation errors, reconciling metrics across different data sources, checking for null values, ensuring date range validity, detecting statistical anomalies, validating metric calculations (median vs mean, rate normalization), checking aggregation grain (per-record vs per-entity), validating contribution analysis for non-additive metrics, or validating consistency across analysis sections. Essential when reviewing analysis before publication, debugging unexpected results, or ensuring data quality in reports. Triggers include "validate this query", "check for errors", "why don't these numbers match", "should I use median or mean", "why don't contributions sum to 100%", "reconcile these metrics", "verify data quality", or any request to catch potential issues in data or calculations.

Discover structural and causal relationships between seemingly unrelated observations to reveal how things influence each other. Use when: (1) asked to find relationships between separate ideas or map how observations connect, (2) multiple observations seem unrelated but might share causal links or common mechanisms, (3) building system models that show interaction patterns and feedback loops, (4) seeking solutions by recognizing analogous mechanisms from different contexts.

Analyze competitor website structures, messaging patterns, and unique features. Use when researching competition, identifying gaps, or informing content strategy.

Proactively learn about the user through onboarding and ongoing observation. Use at session start and when you notice potential preferences.

Marina Moscone competitor-dossier skill. Guides structured competitor research for the MM minisite: prices, lookbooks, e‑comm visuals, press/story arcs, and positioning synthesis. Use when running /mm-comps or doing MM competitor work.

Family recipe collection manager. Extracts recipes from URLs (Instagram, YouTube, websites), images, and Apple Notes. Tracks family preferences, cooking status (To try → Perfected), and syncs with Notion. Supports Hebrew and English. Triggers - "add recipe", "save recipe", "extract recipe", "show recipes", "process recipe inbox", "import from Apple Notes", "sync to Notion", "mark recipe tried", "family recipes", "ninja recipes", "oven recipes".

Complete Medusa backend API reference and integration guide

Extracts high-value insights from research documents, RCAs, design docs, and memos - filters aggressively to return only actionable information. Research equivalent of analyzing-implementations skill.

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes - four-phase framework (root cause investigation, pattern analysis, hypothesis testing, implementation) that ensures understanding before attempting solutions | バグ、テスト失敗、予期しない動作に遭遇した際、修正を提案する前に使用 - 4フェーズフレームワーク（根本原因調査、パターン分析、仮説検証、実装）により、解決策を試す前に理解を確保

Use when analyzing YouTube videos for research, learning, or understanding how content relates to a project - downloads transcripts with yt-dlp, chunks long content, and provides context-aware analysis

Add Wikipedia reference links to concepts that don't have one. Searches for relevant Wikipedia articles and adds them to the references array.

Intelligent command verification for documentation. Discovers all commands in markdown files, validates them using git diff-based cache invalidation, and ensures documentation accuracy with zero token cost after initial setup. Use when asked to verify commands, check documentation, or validate command references.

Enables automatic deep research mode with comprehensive analysis when trigger phrases are detected

Perform preliminary internet research on a person and generate a structured report. Use when the user asks to research, profile, or get information about a specific person by name.

Compare 3D CAD models using boolean operations (IoU, Dice, precision/recall). Use when evaluating generated models against gold references, diffing CAD revisions, or computing similarity metrics for ML training. Triggers on: model diff, compare models, IoU, intersection over union, model similarity, CAD comparison, STEP diff, 3D evaluation, gold reference, generated model, precision recall 3D.

Cross-session persistence for user preferences, implementation guide for Memory feature.

Enforce code style and formatting preferences. Use when writing or reviewing code to ensure consistent style.

Enhanced architecture guidance for voice-first Brazilian fintech applications. Use when designing voice interfaces, implementing PIX/Boletos, optimizing financial systems, or making technology stack decisions for Brazilian market applications. Integrates with docs/ content, MCP tools for Brazilian market research, enhanced validation scripts, and comprehensive Brazilian compliance patterns.

This skill should be used for configuring devloop project settings via .devloop/local.md, git workflow preferences, commit settings, review options

Web research skill for augmenting agent context with current information. Used when outdated information could lead to broken implementations or wasted effort.