Research & Data

Audits an existing service-business homepage or landing page (URL, screenshot, or HTML) for conversion. Returns a 5-second test verdict, then rule-cited findings ranked as quick wins, high-impact changes, and test ideas. Use when a page isn't converting, for a CRO or website review, or when someone shares a URL and asks for feedback. For writing new pages, use website-copy.

Primary pipeline knowledge-capture step that closes the compounding loop. Capture a durable lesson in docs/solutions/ onto the open PR before /closeout merges (the default), or post-merge when the lesson only surfaces during/after merge. Use so future /research and /write-a-prd runs improve. Not for trivial edits with no reusable project-level learning, and not for the merge/teardown itself (that's /closeout).

Use when a design needs references before it needs pixels, and always when `ddr domain match` reports a confidence below what the job needs, a brief lands in new territory, or the work so far looks generic. Runs the expertise loop - maps the best work in the category, adjacent categories and other disciplines, looks at it properly in a real browser, tears it down, hunts the category's slop, writes what it learned into the domain's dossier and reference library, retires what is dated, and re-scores until the agent has earned the right to design.

User-invoked side-route for onboarding to a legacy repo with no docs/solutions/, research archive, or written decisions. Mines the last 12 months of git history and, when gh is available, PR bodies and review threads, into ranked hotspots, cited recovered decisions, and a who-to-ask map, then offers a PR of docs/solutions/ entries. Not for repos that already compound, not for judging people, not model-invoked.

经济学实证研究的动作库——把一次研究拆成八道可执行工序（可行性、识别、数据、估计、表图、报告、写作、审稿），每道工序都有输入、步骤、产出文件与验收标准。Use for any empirical economics research task: feasibility checks, identification design, data cleaning, regression tables, publication-quality figures, reproducible reports, paper writing, and referee responses. Triggers: 怎么做实证、识别策略、数据清洗、回归表、三线表、复现、写论文、回复审稿、研究流程、econ research workflow、empirical pipeline、applied micro.

Use when planning a new Event Packs country pack ("Event Pack: <Country>") or new events for an existing pack: research through subagents, picture search with licence checks, event texts written from sourced fact sheets and passed by the independent editor, and a self-contained READY plan in docs/plans. Ends with a hand-off; event-pack-implement carries the plan out.

Open a pull request that reads like the repo's own maintainers wrote it — pick a depth first, where an express lane for "just open the PR" or "just ship it" orients, writes, and ships without the research passes, and the full lane orients with `ccx vcs lane` and `ccx vcs guidelines`, learns the house style from a bot-filtered sample of recent human-authored merged PRs via a cached per-repo style card, calibrate register against a local wlm voice profile, run a politeness and standards-conformance pass with issue-first handling on repos the user does not own, write the title, body, and commit messages to the repo's own conventions, ship through `ccx vcs ship`, and hand the CI and bot-comment watch to a background pr-watcher agent. Use when opening, submitting, or contributing a pull request, drafting a PR title or body, contributing a change to an open-source repo, or shipping a branch as a PR.

Write a PRD as a durable local artifact at docs/specs/<slug>/_prd.md — research first, clarify with one multiple-choice question at a time, record decisions as ADRs, then write directly without draft-approval rounds.

Create a D3 (Data Driven Decisions) interactive HTML explainer of a research project: one offline file that builds up the concepts, the key derivations, the research gap, and how the code realizes it, with interactive figures (sliders, draggable points, step-through derivations). Written for the repo's own author and co-authors to understand their project better, not for outsiders. Use when: (1) The user wants an explainer, walkthrough, or build-up of a research repo, paper, model, or derivation ("help me understand my own project") (2) Scanning code, paper sub-directories, vault, and results to explain what a project does and why (3) Onboarding a co-author; visualizing a model or derivation interactively (4) Updating an existing explainer after paper or code changed MANDATORY TRIGGERS: explainer, Erklärseite, D3 explainer, HTML explainer, interactive explainer, explain my project, explain my repo, explain the paper, explain the derivation, project walkthrough, co-author onboarding

Primary pipeline tail step after /pre-merge, when a reviewed PR is ready to land and you want the workspace back to a clean base. Use to merge, switch off the worktree before removing it, prune the merged branch, pull base, and verify a clean end state. Triggers on 'close out', 'merge and clean up', 'get back to a clean main/base', 'tear down the worktree'. Not for capturing lessons (that's /compound), or closing the PRD and any other open slices and research-artifact hygiene (that's the cleanup prose in SYSTEM-OVERVIEW Step 9).

Analyze or review a stock portfolio or watchlist with Zacks data, covering return attribution, sector mix, a comparison with SPY and QQQ, and Zacks' view on the largest holdings. Use when the user asks to analyze, review, grade or report on a portfolio, holdings or watchlist.

Work with myFund.pl API v1 — retrieve portfolios and holdings, analyze history, and import transactions, cash flows, and PPK files. Use for myFund.pl integrations or tasks involving data from this API.

Find hidden open-source GitHub repos (MCP servers, skills, subagents, hooks, plugins, libraries) that would make the current task's result better, and security-check them before use. Use when starting a substantial task (build an app, website, video pipeline, scraper, automation, data analysis, etc.), when a <repo-radar> block appears in context, or when the user asks "is there a repo/tool/MCP for X".

Track GST/HST for Canadian freelancers and sole proprietors from a JSON/CSV ledger — per-calendar-quarter GST/HST collected, ITCs, and net tax owing (collected − ITCs) with a T2125-style prep section, the small-supplier alarm (rolling last-4-calendar-quarters taxable income vs the $30,000 registration threshold with exact headroom), and a per-province view (tax system, collected, ITCs, net per province; recorded QST memo figures; PST/RST exposure estimates). Use when preparing GST/HST figures, checking ITCs, monitoring the registration threshold, or seeing place-of-supply exposure province by province. A tracking aid, NOT tax advice and NOT a CRA filing; rates are Nov-2026 defaults, overridable — never present them as live fact.

Select, formulate, implement, compare, or audit portfolio optimization models including MVO, GMV, Black-Litterman, CVaR, CDaR, risk budgeting, HRP, Kelly, robust covariance/mean MVO and Wasserstein DRO. Use for asset allocation, estimation-error control, stress testing, risk decomposition, and point-in-time backtesting.

Native bull/bear/PM debate engine for evaluating a stock (or basket) and producing a decision-grade verdict — the deterministic, Opus-run replacement for the external TradingAgents framework (which is credit-gated). TRIGGER when the user wants a buy/sell/hold call, a "debate this stock", a "what do you think of TICKER", a bull-vs-bear, a valuation sanity-check, to (re)evaluate a SaaS/software name, to re-run a name after earnings, or to refresh/update the Stock Pool hub. Built for software/SaaS but works on any equity. Do NOT trigger for pure data lookups or portfolio rebalancing (use /rebalance).

- Start by naming the job type in one line, such as niche research, competitor research, video analysis, content package or own channel analytics.

Pre-release compliance gate for US digital health software using a two-axis model (legal scope x technical component). Covers what a HIPAA-only review misses - FTC Health Breach Notification Rule, state consumer health data laws, 42 CFR Part 2, ONC information blocking, FDA software rules, tracking pixels, mobile data handling, and AI/LLM data flows. Use when asked to review, audit, gate, or approve a healthcare backend, web app, mobile app, AI feature, or integration before release, or to profi

Turn any research topic, in any field, into a colorful PDF lecture survey of recent papers with figures, research gaps and a table of Q1 paper ideas.

Conducts rigorous entrepreneurial research, validates business and product ideas, investigates real customer demand, competitors, search trends, SEO, marketing, domains, feasibility, monetization and growth, then produces an evidence-audited professional PDF report. Use when evaluating an idea, product, startup, business model, market, technology, brand, domain, go-to-market strategy, or major decision.

Audit how a frontend app talks to its backend and produce a ranked, evidence-backed report of request amplification with concrete fixes - per-row N+1 reads, duplicate fetches, stacked poll timers, zero freshness, blanket invalidation, serial waterfalls, fetch-everything pagination, over-fetching, retry waste. Use this whenever someone asks why an app makes too many requests, wants a network / request / data-fetching / API-call audit, mentions N+1, waterfalls, polling, refetch storms, rate limits

Contrôle de sécurité, RGPD, vente, SEO et qualité avant de publier un site, une boutique, une app ou un SaaS. À utiliser avant toute mise en ligne ou nouvelle version publique.

Validate a product or startup idea end to end and return a research-backed landscape report covering prior art, adjacent incumbents, the graveyard of failed attempts, buyer economics, and the commodity floor. Use this whenever the user asks whether an idea already exists, whether an idea is taken, whether an idea will work, whether anyone has built something, or asks for competitive, prior-art, or market landscape research on something they are thinking of building. Trigger even if they phrase it casually ("has anyone done X", "is this a thing", "am I crazy or is there a market for X") and even if they only describe the idea without asking a question.

Picks the right metrics for a business and explains any metric. Builds a metric pyramid (business goals → unit economics → loyalty → value → quality → traffic sources) with one North Star Metric and its four inputs, industry benchmarks for 31 industries, the three metrics to start with this week and the vanity metrics to skip, and can render it as a page with a PDF for the team. Use whenever the user asks which metrics or KPIs to track, how to measure a business or product, wants a metrics system or dashboard plan, a North Star / NSM, unit economics, LTV, CAC, C1, retention, ARPU, churn, benchmarks or norms for an industry, or asks what a metric means or how to calculate it — e.g. «какие метрики считать», «пирамида метрик», «главная метрика продукта», «что такое C1», «как посчитать LTV», «бенчмарки для кофейни», «на что смотреть каждую неделю». Answers in Russian by default, money in tenge (₸).