research-assistant

name	research-assistant
description	Use for deep, multi-source research that requires synthesis, structured analysis, and iterative evidence gathering before writing the final answer or report.

Research Assistant

Use this skill for substantial research tasks where depth, source quality, and synthesis matter more than speed.

Research Standard

• Do the full job, not the minimum viable version.
• Cross-check important claims.
• Prefer primary and authoritative sources.
• Note uncertainty, missing evidence, and conflicting sources explicitly.
• For active topics, include at least one recency-focused query.

Tooling

Preferred research flow:

• web_search for discovery and breadth
• webfetch or scrape_webpage for source reading
• session_search and session_get for continuity when prior work may exist
• bash for calculations, parsing, or chart generation when needed

For substantial parallel research or long-running build-plus-research work, use the background-session flow from skills/KORTIX-system/sessions/SKILL.md.

Research Continuity

Before deep research on a topic that may have prior history:

• search sessions for prior findings
• reuse or extend earlier work where helpful
• avoid starting from zero if the evidence already exists

Research For Assets

When the task requires both research and a built artifact:

1. Research phase — gather facts, dates, numbers, citations, and source URLs.
2. Asset collection phase — gather verified images, logos, screenshots, or media inputs if needed.
3. Build phase — only then create the website, deck, report, or other asset.

For large parallel workstreams, prefer background sessions over in-turn delegation. Use separate sessions for separate subtopics or regions, then combine the outputs.

Execution Heuristics

• Search snippets are for finding sources, not for trusting facts.
• Read the actual source before treating a claim as established.
• If a ranking or official list exists, go to the original publisher.
• For data-heavy work, inspect intermediate outputs before finalizing conclusions.
• Produce summary artifacts, not just raw gathered material.

Output Quality

Aim for institutional-quality work product:

• readable
• well structured
• evidence-backed
• useful for decision-making

Tables, charts, and comparison matrices are encouraged when they reduce cognitive load.