Research

Mandatory training archive with model gating (APPROVED/REVIEW/DROP). Trigger when: (1) training run completes, (2) need to decide which models to deploy, (3) want historical training reference, (4) need checkpoint recommendations for overfitting.

Design, configure, launch, and analyze ablation sweeps for GRPO training. Use for hypothesis testing, hyperparameter experiments, and systematic comparisons.

Use PROACTIVELY when creating research reports, experiment writeups, technical whitepapers, or empirical study documentation. Orchestrates the complete benchmark report pipeline with structure, diagrams, hi-res PNG capture, and PDF export. Provides working scripts, CSS templates, and complete command sequences for publication-quality AI/ML benchmark reports. Not for slides, blog posts, or simple README files.

Perform SQL joins and multi-table analysis on DST data in DuckDB. Use whenresearch requires combining multiple tables on common dimensions (time, region).Provides patterns for common DST dimension joins and multi-table comparisons.

Naming conventions for planning documents in prompts/. Use when creating plans, PRDs, research reports, idea capture or other workflow documents. Triggers on (1) creating new planning documents, (2) naming PRDs or research reports, (3) questions about document organization in prompts/.

Search RAG database for relevant content. Use for semantic queries over processed documents, code, or papers.

Comprehensive toolkit for creating, analyzing, and visualizing complex networks and graphs in Python. Use when working with network/graph data structures, analyzing relationships between entities, computing graph algorithms (shortest paths, centrality, clustering), detecting communities, generating synthetic networks, or visualizing network topologies. Applicable to social networks, biological networks, transportation systems, citation networks, and any domain involving pairwise relationships.

Research external APIs/services/libraries for feature implementation. Interviews user, researches options, confirms choice, gathers implementation notes. Creates .claude/plans file.

[REPLACE] Teach concepts and demonstrate patterns with examples. Use when [REPLACE with specific triggers].

This skill should be used when the user asks about "MakerKit patterns", "CLAUDE.md files", "what MCP tools are available", "how does MakerKit structure", "personal vs team accounts", "RLS patterns in MakerKit", or needs quick reference for MakerKit conventions.

Automated content publishing pipeline for ID8Labs. Generates essays in Eddie's voice, publishes to id8labs.app, and distributes to social media (X, LinkedIn). Triggers on keywords like release, announce, publish, essay, research article, content pipeline.

Drive quantitative analysis, factor diagnostics, and reporting for ATFT-GAT-FAN outputs.

Manages permanent memory storage for decisions, blockers, context, preferences, and procedures. Use when user says "remember", "save this decision", "what did we decide", "recall", "search memories", "any blockers", or when making important architectural decisions. Provides SDAM compensation through external memory.

Systematic debugging methodology that eliminates guessing and speculation. Add instrumentation to gather specific data that fully explains the problem. Evidence before hypothesis. Observation before solution.

Extracts C++ code snippets, lists classes and functions, and provides source code references for comprehensive C++ codebase analysis. This skill should be used when analyzing any C++ source files to understand code structure, extract reusable components, and generate structured documentation.

Web and code search with Exa MCP. Use for current documentation, API references, code examples, latest library info, or when the user mentions "exa", "web search", "docs", or "current API".

Structured code review output format for the reviewer agent. Provides APPROVED/ISSUES format with file:line references. Use when: reviewing code changes, code quality checks, architecture compliance checks, approving code for merge, identifying critical/major/minor issues.

Draft a Conventional Commits message from either a short summary of changes or by inspecting the local git diff. Use when the user wants a commit message formatted as type(scope): subject, with optional body and footers for breaking changes or issue references.

Invoke Gemini CLI as a partner AI developer for tasks benefiting from its 1M token context window or a second opinion. Use when analyzing large codebases, generating documentation, getting alternative perspectives, researching topics with web search, or creating UI wireframes/components. Gemini has read/write/edit/shell/web search capabilities like Claude.

Analyze GRPO training runs for learning dynamics and pipeline performance. Use when diagnosing training issues, reviewing Elo progression, checking throughput, or updating experiment results.