Research skills

AI skills
‍for researchers

Give your AI the skills to work like a researcher.

Not a collection of prompts. Each skill encodes an accepted research methodology: it asks the right questions before it starts, analyses evidence correctly, knows when it should not be used, and states its own limitations.

Built by researchers. Free to use. Vendor-neutral, and it works on Claude, ChatGPT, Gemini or any assistant.

107 skills

15 categories

5 shared protocols

Try:“analyse interview transcripts”·“clean survey data”·“build a questionnaire”·“find insights”

Search results

Results for

Clear search

We couldn't find an exact match.

Try one of these related research areas:

Can't find the skill you need? Suggest a skill →

Undergraduate to doctorate, supervision and examination, and academic publishing.

28 Skills

Browse →

Extracts what a brief is actually asking: the decision behind it, the assumptions inside it, the scope boundary, and the questions that must be answered first.

Works withThe brief exactly as written, in the requester's own words, Provenance: who wrote it, who commissioned it, when, and why, Access to someone who can answer questions, or a statement that access is closed, The email or meeting thread the brief came out of (optional), Previous briefs and previous research from the same requester (optional), Stakeholder list with roles (optional), Budget order of magnitude and decision date (optional)

AvailableView skill →

research-strategy-and-designresearch brief, brief interrogation, clarifying questions research, what is the client actually asking, research intake, stakeholder brief, research request triage, scope boundary, out of scope research, unstated assumptions in a brief, research objectives clarification, method before question, confirmation bias in research briefs, ratification research, research kick off questions, client brief review, research proposal preparation, stakeholder mapping researchBrief Interrogation Note, Assumption register with consequence and verification action per line, Decision statement, or an explicit "decision not established", Method-before-question finding, Stated, inferred and anti success criteria, Stakeholder map with the decision owner identified, Three-part scope boundary including silent inclusions, Constraints expressed as designs already ruled out, Confirmation-seeking assessment, Blocking clarifying questions, separated from non-blocking ones with defaults, One-page restatement for the requester to correct

01.02Research Strategy

Works withPrepared dataset with cleaning, missing-data and transformation logs, Achieved sample composition and the quota or intended composition, Precise universe definition, Target figures with source, date and the population they describe, Sample design and sample type, Previous waves' weighting specifications and efficiencies (optional), Design weights or selection probabilities (optional), Auxiliary variables correlated with the key measures (optional), Analysis plan and reporting structure (optional)

AvailableView skill →

data-preparationsurvey weighting, rim weighting, cell weighting, post-stratification, design weights, weighting targets, weight efficiency, effective base size, effective sample size, weight capping, extreme weights, weighted percentages, unweighted counts, base reporting, subgroup bases, non-probability sample, coverage error, non-response bias, representativeness, tracker weighting, weighting disclosureWeight variable, or a documented decision not to weight, Weighting disclosure block naming what the weighting does not correct, Composition table with achieved, target, source, date and post-weight figures, Weight diagnostics including range, mean, extremes and who carries them, Weight efficiency and effective base for the total and every reported subgroup, Capping comparison with effective base, target deviation and key measures, Weighted versus unweighted headline comparison, Reject-or-weight judgement with its evidence, Versioned scheme specification for trackers, Residual limitation statement

05.01Quantitative Analysis

Works withThe call for submissions, with tracks, word limit, format and selection criteria, The work at whatever stage it has actually reached, stated honestly, The format sought (talk, poster, lightning session, panel, symposium), Optional previous conference programme, Slot length and room format, Audience composition, Optional existing paper as source material

AvailableView skill →

academic-university-researchacademic-publishingconference abstract, structured abstract, conference submission, abstract rejected, research poster design, conference presentation, lightning talk, one idea per slide, talk timing and overrunning, question session preparation, work in progress track, academic conference networking, presenting unfinished researchSubmission package mapping each selection criterion to a sentence in the abstract, Structured or prose abstract with all four contents present, Completeness statement covering what will exist by the conference date, Presentation design table (one message, supporting points, time budget, cut list), Slides or poster built to distance-readability discipline, Ten prepared question answers, including the feared ones, Conference plan with named people, sessions and a follow-up deadline

15.28Academic Research

Grant and Funding Proposal Development

Write to the call's assessment criteria, argue feasibility and impact instead of asserting them, and plan around a base rate most applicants are never told.

Works withThe full call document, including assessment criteria and hard rules, A research idea at the level of a question, aim and approach, The team's actual track record, supplied by the people concerned, Optional funder strategy, priorities and previously funded projects, Optional feedback and scores from a previous unsuccessful application, Collaborator and partner commitments, agreed or in negotiation, Costing information from the institution's finance function, Optional pilot or preliminary data

AvailableView skill →

academic-university-researchacademic-publishinggrant application, funding proposal, case for support, assessment criteria, funder priorities, work plan and milestones, risk register, budget justification, pathways to impact, data management plan, research ethics section, grant resubmission, funding success rates, fellowship application, lay summaryCriteria map with weight, location, evidence and owner per criterion, Hard-rules checklist covering eligibility, limits, sections and deadlines, Funder fit statement naming the priorities served and how, Case for support with the research question stated for specialist and lay readers, Work plan with dependencies, milestones and slack, Risk register including the scientific risk, Team justification mapping capability to work package and committed time, Budget mapped line by line to the work plan, with a structural justification, Impact plan with beneficiaries, pathway, costed activities and indicators, Data management, ethics and governance sections, Resubmission statement and portfolio plan

Five protocols every skill inherits.

This is the difference between a skill library and a prompt pack. The protocols are what stop AI-assisted research going wrong, and they are not optional.

Skill Router

Matches your situation to the right skill, and tells you what comes next. In a library this size the skill you most need is often the one you did not know existed.

Evidence Traceability Protocol

Every conclusion stays traceable back to the source it came from, so a reader can check any claim rather than take it on trust.

Confidence and Uncertainty Protocol

Say strong things strongly and weak things weakly. Confidence in the output has to match the evidence underneath it.

AI Guardrail Standard

The prohibitions that never vary. No invented data, respondents, quotes or sources. No significance without a test. No correlation reported as cause.

Human-in-the-Loop Standard

Where a researcher's judgement is required and the AI must stop: business relevance, cultural interpretation, high-stakes recommendations and ambiguous qualitative evidence.

What these skills will not do

They will not invent data, respondents, quotes or sources. They will not claim statistical significance without a test. They will not turn correlation into causation. They will not hide evidence that contradicts the conclusion. They will not make a recommendation stronger than the evidence supports. Where the evidence is insufficient, they say so and stop. That is the point of them.

How it works

Add research expertise to your AI.

Choose a skill

Find the research task you need help with, by problem, by category or by where you are in the project.

Load it with the kernel

Add the skill file and the five protocols to a Claude Project, a ChatGPT Project, a Gemini Gem, or paste them into any assistant.

Give it the real material

Your actual brief, questionnaire, transcripts or dataset, not a description of them. Then answer its pre-start questions honestly, including "I do not know".

Your material

Brief, questionnaire, transcripts, dataset

Skill plus kernel

An encoded method, and the protocols that keep it honest

What you get back

Structured, evidenced, traceable, with its limits stated

One library. The whole research lifecycle.

From a business problem arriving to insights being reused a year later. Every skill declares what comes before it and what comes next.

Run against any of it, at any point

AI Output Verification Bias Detection Research Quality Review Literature Review and Desk Research

Neutrality

Useful, or nothing.

Every skill in this library is vendor-neutral. No file names or recommends any research platform, panel, product or supplier, and no skill routes a researcher toward a commercial outcome. Method Selection recommends the method the research question requires, including methods no supplier can execute, and including the recommendation not to run primary research at all where the answer already exists.

The library is built to augment professional researchers, not to pretend they are unnecessary. Business relevance, cultural interpretation, strategic implication, ethical appropriateness and high-stakes recommendations all return to a person. K5 sets out exactly where and why.

Built for researchers. By researchers.

Research methods are changing quickly, and this library will keep growing with them. If a skill you need is missing, tell us and it goes on the register.

Browse all 107 skills Suggest a skill

Yazi is Zulu for to know.

Research where people already are.
‍Analyse it where you already work.

Yazi helps researchers conduct surveys, AI interviews and longitudinal research directly through WhatsApp.

Explore Yazi Book a demo

AI skills ‍for researchers

Results for

We couldn't find an exact match.

What do you need help with?

Research Strategy and Design

Instrument Design

Fieldwork and Data Collection

Data Preparation

Quantitative Analysis

Specialist and Advanced Analysis

Qualitative Analysis

Insight Development

Segmentation and Audience Understanding

Desk Research and Evidence Synthesis

Reporting and Storytelling

Report Design and Compilation

Research Quality, Ethics and Governance

Activation and Knowledge Management

Academic University Research

Built first, built deepest

Business Problem to Research Question

Survey Questionnaire Design

Descriptive Analysis

Thematic Analysis

Finding to Insight Development

Research Report Compilation

Explore all skills.

Research Brief Interrogation

Business Problem to Research Question

Hypothesis Development

Research Method Selection

Research Plan Development

Sampling Strategy

Analysis Plan Development

Research Proposal and Scope Development

Survey Questionnaire Design

Discussion Guide Design

Interview Question Development

Question Bias Detection

Survey Logic and Flow Review

Screener and Quota Design

Scale and Measurement Selection

Recruitment and Sample Sourcing

AI-Moderated Interview Design

Stimulus and Concept Preparation

Fieldwork Monitoring and Response Quality

Incentive and Participation Design

Data Validation

Data Cleaning

Missing Data Handling

Data Transformation and Dataset Preparation

Weighting and Base Management

Descriptive Analysis

Statistical Testing

Cross-Tabulation

Driver Analysis

Trend and Tracker Analysis

Correlation, Regression and Causal Claim Control

Conjoint and MaxDiff Analysis

Pricing Research Analysis

Brand Health and Equity Analysis

Experiment and A/B Test Analysis

Customer Experience and Journey Analysis

Large-Scale Text Analytics

Thematic Analysis

Open-Ended Response Coding

Interview and Transcript Analysis

Quote and Evidence Extraction

Sentiment and Emotion Analysis

Qualitative and Quantitative Integration

Finding to Insight Development

Pattern Identification

Implication Development

Recommendation Development

Insight Prioritisation and Sizing

Audience Segmentation

Persona Development

Behavioural Profiling

Segment Comparison

Needs, Motivation and Jobs-to-be-Done Analysis

AI skills
‍for researchers