Iterative Research Loop Skill
Run structured research loops that search, compare, verify, refine, and synthesize findings with clear stop conditions, source checks, and final reports.
Iterative Research Loop Skill is a free, reviewed AI skill for agent systems & llm workflows. Run structured research loops that search, compare, verify, refine, and synthesize findings with clear stop conditions, source checks, and final reports. It works with ChatGPT, Claude, Gemini and is ready to use out of the box.
- • The skill cannot verify current facts, prices, rankings, or market data without live browsing or reliable source access.
- • It should not invent sources, citations, statistics, quotes, dates, authors, or findings.
- • Research quality depends on the sources and context provided.
About this skill
Iterative Research Loop helps users get more reliable answers than a single first pass by guiding an AI assistant through a controlled, multi-step research process. It defines the research question, breaks it into sub-questions, creates search angles, gathers and compares sources, identifies gaps and contradictions, refines the search only as needed, and produces a final synthesis with clear stop conditions rather than running indefinitely. It is useful for market research, technical research, competitor research, academic-style exploration, product discovery, SEO research, tool comparison, trend research, and decision reports. The skill is careful never to invent sources, statistics, quotes, or findings, and clearly flags when live browsing or verification is required before a claim can be trusted. This makes it especially valuable for people who need decision-ready research output rather than a single unverified answer, while staying honest about the limits of what can be confirmed in a given session.
What it does
This skill guides an AI assistant through a structured research loop: defining the question, decomposing it into sub-questions, selecting search angles, gathering and evaluating sources, detecting gaps and contradictions, refining the search only when needed, and producing a final synthesis that separates confirmed findings, likely findings, assumptions, and unknowns, all within a defined loop limit and clear stop conditions.
What is included
- SKILL.md — concise runtime instructions for the AI assistant
- workflow.md — step-by-step workflow for running controlled iterative research loops
- research-loop-framework.md — framework for research phases, loop limits, source evaluation, gap detection, contradictions, and synthesis
- source-quality-checklist.md — checklist for evaluating authority, recency, relevance, bias, evidence quality, and confidence
- gap-and-contradiction-checklist.md — checklist for finding missing information, verifying claims, handling contradictions, and deciding whether another loop is needed
- output-templates.md — reusable formats for research plans, loop logs, source tables, gap analysis, contradiction reports, and final reports
- examples.md — realistic input and output examples for iterative research use cases
How to use it
1. Download the ZIP file for this skill 2. Extract the files to a folder on your computer 3. Open your AI assistant or research assistant 4. Upload or paste the skill files if your tool supports custom skills or knowledge files 5. Provide the research question, decision context, desired depth, target market, timeframe, and output format 6. Tell the assistant whether live browsing or research tools are available 7. Ask the assistant to apply the Iterative Research Loop Skill
Examples
I want to research whether an AI skills library like PiSkill has demand. Please use a structured research loop. Focus on AI builders, prompt users, no-code builders, and people using tools like Claude, ChatGPT, Cursor, Lovable, and Codex. Do not invent market numbers.
Research loop plan: The research should examine demand signals from existing AI prompt libraries, skill marketplaces, no-code builder communities, search behavior, competitor positioning, and user pain points. Loop 1 should map the market and identify existing resource types. Loop 2 should compare competitors and look for gaps such as quality control, safety, downloadable skill files, and request-based creation. Loop 3 should verify whether users complain about generic prompts, unsafe uploads, poor organization, or lack of practical examples. Final output should separate confirmed findings, likely opportunities, assumptions, and questions that require live validation.
Known limitations
- The skill cannot verify current facts, prices, rankings, or market data without live browsing or reliable source access. - It should not invent sources, citations, statistics, quotes, dates, authors, or findings. - Research quality depends on the sources and context provided. - It cannot replace expert review for legal, medical, financial, engineering, or high-stakes research. - Loop-based research must remain controlled with clear stop conditions and should not run endlessly.