Systematic research workflows for SOTA AI agents. Includes tool-triggered searches, multi-source synthesis, source credibility evaluation, bias detection, structured reporting, and citation practices.
NewMaintained
Automatic
◆Research pack
Last verified 8/9/2026
What it changes
4 of 22 test cases passed with the skill.
Pass-rate lift
—
Tokens
+59%
Agent turns
0%
Response time
+11%
Model gemini-3.6-flashMeasured 8/9/2026v14 of 22 cases passed with skill · 4 comparable
Tokens, turns and time describe resource use, not a dollar price.
Lift withheld: 4 comparable cases; this report requires 8.
See it in action
A test case the skill fixed. Same model, same prompt, with and without the skill.
Prompt
Please summarize the following text into three bullet points: 'The company reported a 15% increase in Q3 revenue reaching $4.2 billion, driven primarily by enterprise cloud adoption. Operating margins expanded to 28% from 24% in the prior year quarter. Executive leadership raised full-year guidance by $200 million.'
What was checked, without and with the skill
Checked
Without skill
With skill
The response directly summarizes the provided text into three concise bullet points.
Missed
Passed
The response does not trigger external web searches.
Passed
Passed
case-19 · gemini-3.6-flash · 8/9/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md36 linessingle file
Research Assistant
When active, guide the agent through rigorous, transparent research processes.
Core Principles
Always use tools explicitly for current information (web search, X search, browse specific pages).
Cross-verify information across multiple independent sources.
Distinguish between facts, opinions, and speculation.
Evaluate source credibility, recency, and potential bias.
Structure outputs clearly with sources and confidence levels.
Recommended Workflow
Clarify Scope: Define exact research questions and success criteria with the user.
Tool Activation: Explicitly trigger searches: "Search the web for...", "Search X for recent discussions on...", "Browse [specific authoritative page]".
Gather & Synthesize: Collect key points from multiple sources. Note agreements, contradictions, and gaps.
Evaluate: Assess reliability of each source. Flag potential biases or outdated info.
Structure Output:
Executive Summary
Key Findings (with sources)
Conflicting Views / Uncertainties
Recommendations or Next Steps
Full Source List with dates/access info
Iterate: Offer to dive deeper into specific areas or update with new searches.
Output Quality Rules
Every claim should be traceable to a source.
Use tables for comparisons when helpful.
Include confidence levels or "as of [date]" notes for time-sensitive info.
Be transparent about limitations of available data.
Pitfalls to Avoid
Relying on single source or model training data alone.
Presenting speculation as fact.
Over-generalizing from limited results.
This skill produces reliable, well-sourced research outputs.