research

Use when starting a scaffolded research workspace that needs an explicit project profile, methodology, audience-output matrix, wave roadmap, and dispatch gate; do not use for routine work inside an already initialised project.

Use when an academic paper, thesis, dissertation, journal article, systematic review, or examinable artefact must follow a named reporting or examination standard; use academic-writing instead for prose craft without a reporting-standard decision.

Use when drafting or revising an academic paper, essay, thesis, or dissertation with source-away composition, originality, citation, synthesis, voice, and register controls; use academic-reporting-standards for selecting formal study-reporting rules.

Use when designing, running, or auditing multi-agent research operations: wave planning, task briefs, agent roles, tool boundaries, verification loops, failure recovery, evaluation, and portable runtime behavior across Codex, Claude, and other agents.

Use when designing evaluation, failure analysis, feedback capture, regression datasets, synthetic data controls, and data flywheels for AI-assisted research and agentic workflows.

Use when auditing a concrete artefact for AI slop after a major iteration or before release, with evidence-backed findings, fixes, genericness score, and A/B/C/F verdict; use anti-ai-slop during production rather than for retrospective grading.

Use when a forward-looking judgment or contested-evidence question needs structured analytic techniques, bias checks, sourcing discipline, alternatives, and calibrated estimative language; use critical-reasoning-and-argument for general claim and warrant testing.

Use when selecting or designing the exact shape of a decision-grade analytical artifact: intelligence brief, estimative memo, warning note, decision memo, status report, evaluation report, research report, due-diligence memo, market analysis, or executive one-pager.

Use when producing or revising any research artefact to prevent generic, unsupported, repetitive, or mechanically styled output in real time; use ai-slop-audit to grade a concrete completed iteration independently.

Use when drafting or revising non-academic prose such as emails, memos, letters, plans, articles, web copy, speeches, resumes, or executive briefs; use academic-writing for scholarly outputs and executive-communication for decision-led executive synthesis.

Use when research output makes forecasts, probability judgments, risk calls, market outlooks, scenario likelihoods, warning indicators, or confidence claims that must be calibrated, scored, updated, and separated from source reliability.

Use when starting a new project, scoping a feature, or deciding which skill stack to load — provides the canonical Foundation → Implementation → Validation → Companions skill set per technology domain. The lookup table that turns "I'm building X" into "load these skills, in this order".

Use when the engine's research must be turned into consulting-grade problem solving: issue trees, hypothesis-led workplans, analysis plans, stakeholder management, implementation logic, and transformation framing. Encodes consulting craft without relaxing evidence discipline.

Use when an output makes claims, explanations, interpretations, judgments, recommendations, forecasts, or persuasive conclusions that need explicit warrants, assumptions, countercases, and implications; use analytic-tradecraft for estimative or contested-evidence methods.

Use when profiling, cleaning, merging, or validating tabular research data through encoding, tidy-structure, anomaly, lineage, and quality gates; use dataset-discovery-and-analysis first when the dataset still needs to be found or retrieved.

Use whenever a research project would benefit from public datasets — finds them across government / IGO / academic / ML hubs, retrieves with integrity checks, and produces a Segnini-style five-step quality profile (completeness, duplicates, accuracy, integrity, codes). Backed by tools/datasets/.

Use when research must support a decision, recommendation, investment, policy choice, product bet, go/no-go call, prioritization, or executive memo. Encodes decision framing, alternatives, values, uncertainty, tradeoffs, pre-mortem, and commitment quality gates.

Use when planning, drafting, reviewing, or formatting a dissertation from research question through defence-ready manuscript; use academic-writing for prose originality and research-design for method selection.

Use when generating or revising a project's triple-layer AGENTS.md documentation from observed workspace structure, technology, data, and planning evidence; do not use for general technical documentation or a single ad hoc README edit.

Use when defining or auditing the mandatory operating sequence for the research engine: evidence discipline, research design, collection, verification, reasoning, tradecraft, calibration, decision support, report shape, productization, and export.

Use when conducting lawful corporate, financial, sanctions, regulatory, ownership, or background-check due diligence that needs an auditable finding trail and CARA resolution; use osint-investigation for broader open-source inquiry without a diligence decision.

Use when English content must sound natural and professional for East African readers; apply British-influenced regional usage after the content-owning skill, but route translation or substantive content strategy elsewhere.

Use when research must become a traceable evidence-to-claim knowledge graph rather than loose notes. Encodes source, evidence, claim, warrant, inference, synthesis, gap, contradiction, and finding nodes with registry-ready relationships.

Use when creating, importing, repairing, or validating Excel workbooks with formulas, tables, charts, controls, and professional formatting; use python-document-generation for multi-format application exports.

Use when verified research must become an answer-first memo, report, deck, or one-pager for senior readers; this skill structures the argument and page messages, while critical-reasoning-and-argument tests the underlying claims.

Use when an academic proposal, thesis, dissertation, project, defence, or research-governance artefact must be checked against a named Kenya-based handbook; use primary-research for fieldwork design and source-evaluation for evidence quality.

Use when turning a corpus into reusable, monetizable, verifiable knowledge assets: entity-event-issue extraction, claim libraries, evidence maps, reusable briefs, ontology seeds, dossiers, and refresh cadences.

Use when the engine must turn research into reusable knowledge assets, audience-specific variants, and monetizable offerings rather than a one-off document only. Encodes knowledge audit, asset design, reuse discipline, and commercialization framing.

Use when creating or updating an end-user manual or reference guide for an ERP or SaaS feature; use project-requirements for product scope and professional-word-output for final DOCX production.

Use when Markdown files need mechanical lint repair for headings, list spacing, fences, and whitespace without changing meaning; use a writing or editing skill when prose or structure needs substantive revision.

Use when a verified corpus or brief must be compressed into a navigable mind map for synthesis, taxonomy, wave planning, or decomposition; use cross-cohort-synthesis for final comparative findings and knowledge-mining for reusable claim assets.

Use when a question turns on statutes, regulations, cases, treaties, legal authority, or jurisdiction-specific analysis; use source-evaluation for general source quality and refuse to treat unverified currency or secondary commentary as controlling law.

Use when lawful public-source investigation needs stakeholder reconnaissance, adverse-media review, chronology, source capture, or case-vault discipline; use pi-investigation only for authorised licensed-PI work and refuse surveillance, harassment, stalking, or doxxing.

Use when research, analysis, forecasts, or recommendations need adversarial review, red-team challenge, devil's advocacy, source audit, method review, dissent handling, and revision disposition before release.

Use when an authorised licensed private-investigator workflow needs lawful collection, chain of custody, and an evidentiary report; use osint-investigation for public-source reconnaissance and refuse unlicensed surveillance or harm-enabling requests.

Use when the project cannot rely on desk research alone and needs first-hand qualitative evidence such as interviews, observation, focus groups, field visits, or coded open-ended responses. Encodes qualitative design, fieldwork discipline, coding, credibility checks, and reporting gates.

Use when producing or quality-checking a polished Microsoft Word document with controlled styles, pagination, tables, figures, and render evidence; use manual-guide for end-user content and python-document-generation for automated multi-format exports.

Use when a guided discovery interview must produce requirements, business rules, user types, and workflows for a new SaaS project; use systems-process-requirements when formal traceability, interfaces, states, and acceptance criteria are the primary need.

Use when Python must generate downloadable Excel, Word, or PDF artefacts for application users with branding, charts, tables, formulas, and delivery controls; use the format-specific skill when only one document type needs manual production.

Use when research requires quantitative modelling: market sizing, triangulation, sensitivity analysis, scenario models, assumption registers, forecast models, data-quality caveats, and model audit checks.

Use when choosing and drafting a long-form business report, proposal, business plan, bid response, or white paper whose evidence, persuasion structure, and reader decision must be explicit; use business-writing for short emails, memos, blogs, or web copy.

Use when a research project needs formal questions, method selection, case logic, historical or trend design, MROC design, knowledge-lifecycle modelling, or a research-design document; use research-orchestration for wave planning and research-techniques for a single named technique.

Use when starting or replanning a multi-source research project that needs a research-type choice, discipline strategy, reading mode, wave plan, verification, gap filling, and synthesis; use research-design for formal method design and research-techniques for one bounded technique.

Use when selecting the document form and academic or non-academic variant for a research deliverable such as a paper, thesis, policy brief, intelligence report, market analysis, legal opinion, case study, or feasibility study; use report-and-proposal-craft for the persuasion mechanics of business reports and proposals.

Use when a research project needs a specific named technique — mini-analysis, crosswalk matrix, cross-cohort synthesis, gap analysis, reference interview, controlled vocabulary, pain-point taxonomy, search-operator grammar, Google-search API. Companion to research-orchestration; load only the technique the situation demands.

Use when scaling a Python scraper beyond a single-file requests script — encodes Lawson's *Web Scraping with Python* engineering layer (caching to disk and MongoDB, concurrent downloading, dynamic content via WebKit/Selenium/Playwright, form interaction, CAPTCHA strategies, Scrapy framework selection).

Use when defining or reviewing how multiple skills compose through shared house style, input/output artifacts, evidence declarations, and reference boundaries; use skill-writing for one skill's authoring workflow and skill-taxonomy-and-routing for catalogue placement.

Use when performing a read-only safety review of a new or changed skill for credential harvesting, unsafe installers, hidden mutation, excessive permissions, unknown tools, or instruction conflicts; use skill-writing for structural quality and source-verification for factual claims.

Use when normalizing skill names, resolving whether a capability should be a standalone skill or a reference file, updating routers, removing alias drift, or auditing the repository's research skill taxonomy.

Use when creating or upgrading a skill in this research engine with portable frontmatter, progressive disclosure, reference-file strategy and validation under the canonical chwezi-dev-engine skill-writing standard; use skill-composition-standards to normalise house style and skill-safety-audit for read-only review.