Research

Evaluate and elevate writing effectiveness through multi-dimensional quality assessment. Analyzes craft, coherence, authority, purpose, and voice with genre-calibrated thresholds. Use for refining drafts, diagnosing quality issues, generating quality content, or teaching writing principles.

Web research via Perplexity AI. Use for technical comparisons (X vs Y), best practices, industry standards, documentation. Triggers on "research", "compare", "vs", "best practice", "which is better", "pros and cons".

Use when writing or reviewing NIH, NSF, or foundation grant proposals. Invoke when user mentions specific aims, R01, R21, K-series, significance, innovation, approach section, grant writing, proposal review, research strategy, or needs help with fundable hypothesis, reviewer-friendly structure, or compliance with grant guidelines.

Use when reviewing or editing research manuscripts, journal articles, reviews, or perspectives. Invoke when user mentions manuscript, paper draft, article, research writing, journal submission, reviewer feedback, or needs to improve scientific writing clarity, structure, or argumentation in their manuscript.

Use Gemini CLI for research with Google Search grounding and 1M token context

Research-to-release SOP for high-conversion landing pages with validated copy, design, and deployment steps.

Advanced defect eradication with hypothesis-driven RCA, guarded fixes, and comprehensive validation.

Reference the LOCAL connect-go/ directory for ConnectRPC protocol. NEVER use WebFetch/WebSearch for github.com/connectrpc/connect-go - always read local files.

Use when you need to generate many creative options before systematically narrowing to the best choices. Invoke when exploring product ideas, solving open-ended problems, generating strategic alternatives, developing research questions, designing experiments, or when you need both breadth (many ideas) and rigor (principled selection). Use when user mentions brainstorming, ideation, divergent thinking, generating options, or evaluating alternatives.

Use when investigating why something happened and need to distinguish correlation from causation, identify root causes vs symptoms, test competing hypotheses, control for confounding variables, or design experiments to validate causal claims. Invoke when debugging systems, analyzing failures, researching health outcomes, evaluating policy impacts, or when user mentions root cause, causal chain, confounding, spurious correlation, or asks "why did this really happen?"

Research-driven essay and post creation with thematic synthesis, citation management, and voice calibration. Use when creating Substack/LinkedIn posts, long-form essays synthesizing multiple sources, or publication-grade writing requiring web search, narrative arc, and proper attribution. Triggers include "research and write about [topic]" or "dig into this idea and write."

Implements tasks from .plans/ directories by following implementation guidance, writing code and tests, and updating task status. Use when task file is in implementation/ directory and requires code implementation with comprehensive testing. Launches research agents when stuck.

Guide when to use built-in tools (WebFetch, WebSearch) and MCP servers (Parallel Search, Perplexity, Context7) for research. Synthesize findings into narrative for braindump. Use when gathering data, examples, or citations for blog posts.

Establish observability for research systems, experiments, and data pipelines with guardrails and confidence ceilings.

Generates, explains, and plans shell commands, bash scripts, unix pipelines, awk/sed/grep expressions. Use when user requests shell commands, bash scripts, terminal commands, unix pipes, scripting help, or command syntax assistance. Do not use for web research or document analysis.

Execute focused research tasks end-to-end with clear constraints, evidence tracking, and confidence ceilings.

Query Google NotebookLM for source-grounded, citation-backed answers from uploaded documents. Reduces hallucinations through Gemini's document-only responses. Browser automation with library management and persistent authentication.

Set up complete experimental infrastructure for all runs in a designed experiment. Orchestrates parallel generation of fine-tuning configs (via scaffold-torchtune) and evaluation configs (via scaffold-inspect). Use after design-experiment to prepare configs before running experiments.

Design and iterate on new research methods with structured checkpoints, baselines, and validation.

Research and extract an engineer's coding style, patterns, and best practices from their GitHub contributions. Creates structured knowledge base for replicating their expertise.