convez.ai
SkillDesenvolvimento WebAvançado
lyndonkl

Claude Code Skills Collection

lyndonkl/claude

241 skills + 62 agentes para análise, escrita, design e ML em Claude

Ver no GitHub
BomPluginScore · de 0 a 100

Biblioteca pronta para produção com 241 skills e 62 agentes orquestrados para Claude, cobrindo frameworks de pensamento, pesquisa, escrita, design, dados/ML, finanças corporativas e análise estratégica. Instalação em 30 segundos via plugin marketplace, com maioria dos skills funcionando out-of-the-box.

Instalação

$npx convez-ai lyndonkl/claude

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
200
Forks
25
Acompanhando
1
Saúde
28%
Último commit
há 2 meses
Versão
0.1
Issues abertas
0
Contribuidores
2

Casos de uso

  • Análise financeira e valuação de empresas (M&A, IPO)
  • Pesquisa macroeconômica e thematic analysis
  • Escrita técnica, científica e conteúdo (papers, blogs, grants)
  • Design de dashboards e visualizações cognitivas
  • Construção de sistemas GraphRAG e knowledge graphs
  • Gestão financeira pessoal e corporate finance

Veredito

Repositório maduro e bem documentado com cobertura ampla de casos de uso (finanças, análise, ML, escrita). A arquitetura de agentes + skills oferece composição poderosa, mas requer familiaridade com Claude Code e a filosofia de agentes orquestrados. Excelente para quem já domina ferramentas de IA e busca extensibilidade séria.

README

A tradução para o português deste README ainda está na fila.

Claude Code Skills Collection

Skills Agents Status Run in Smithery

A production-ready library of 241 skills and 62 orchestrating agents for Claude Code — covering thinking frameworks, research, writing, design, data/ML, corporate finance, game theory, fantasy baseball, an evolutionary 10-agent FIFA World Cup fantasy backroom, household personal finance, a 9-agent team for growing a Substack publication, and a 9-agent learning studio for becoming a contributor to ML-driven crop genetics / genomic selection.

Install in 30 seconds:

/plugin marketplace add lyndonkl/claude
/plugin install thinking-frameworks-skills

Note: A small number of skills wrap native CLI tools (currently only markdown-to-pdf, which needs pandoc and a LaTeX engine). Those tools must be installed separately. See Optional native dependencies below. All other skills work out of the box.


I want to…

Pick the fastest entry point for what you're trying to do. Most users start with an agent (which orchestrates many skills); power users invoke skills directly.

I want to…Start here
Value a company / analyze an M&A target / plan an IPOcompany-analyst, acquisition-analyst, ipo-strategist, special-situations-analyst
Research a thematic macro trend cluster (AI compute cycle, demographic shift, climate transition, etc.)macroeconomic-analyst
Reverse-engineer a real product's vision / strategy / tactics / ML and system architecture from public materialproduct-strategist
Decide how to allocate capital (debt / dividends / projects)capital-allocation-strategist
Manage my Yahoo Fantasy Baseball teammlb-fantasy-coach + 6 MLB specialists
Run my FIFA World Cup Fantasy squad (evolutionary boards — agents propose, you decide)wc-director + 9 team agents
Run my household finances from PDF statements (drop in, briefing + dashboard out)household-cfo + 8 household specialists
Write a paper / grant / recommendation letterscientific-writing-editor
Improve any piece of writing (blog, memo, essay)writing-assistant
Make a calibrated forecast or probability estimatesuperforecaster
Design a dashboard, viz, or UI grounded in cognitioncognitive-design-architect
Build a GraphRAG / knowledge-graph retrieval systemgraphrag-specialist
Design equivariant / symmetry-aware neural networksgeometric-deep-learning-architect
Build geometric intuition for ML math (attention, PCA, eigenvectors, high-dim spaces)math-intuition-coach
Get a weekly digest of new bioRxiv / medRxiv / PubMed / arXiv papers (life sciences + CS / ML) against my keyword watchlistliterature-scan-coach → paper-synthesizer
Grow my Substack / publish intuition-first ML & systems essayslibrarian, intuition-builder, editor + 6 more
Learn ML-driven crop genetics / genomic selection (experiential study → notes → publish in public)biostat-tutor, biostat-assessor, biostat-editor + 6 more
Use just one tool (no agent)Browse the Skills Index below

How the pieces fit together

flowchart LR
    U([What do you need?]) --> D{Task}
    D -->|Analyze a company| CA[company-analyst<br/>acquisition-analyst<br/>ipo-strategist]
    D -->|Analyze a product| PS[product-strategist]
    D -->|Write something| WA[writing-assistant<br/>scientific-writing-editor]
    D -->|Forecast / decide| SF[superforecaster]
    D -->|Design / visualize| CD[cognitive-design-architect]
    D -->|Fantasy baseball| MLB[mlb-fantasy-coach<br/>+ 6 specialists]
    D -->|World Cup fantasy| WC[wc-director<br/>+ 9 team agents]
    D -->|Household finances| HF[household-cfo<br/>+ 8 specialists]
    D -->|Knowledge retrieval| GR[graphrag-specialist]
    D -->|Symmetry-aware ML| GDL[geometric-deep-<br/>learning-architect]
    D -->|Math intuition| MI[math-intuition-<br/>coach]
    D -->|Weekly paper digest| LSC[literature-scan-coach<br/>→ paper-synthesizer]
    D -->|Learn crop genomics| BIO[biostat-tutor<br/>+ 8 specialists]
    D -->|One-off tool| SK[Skills Index ▾]
    CA & PS & WA & SF & CD & MLB & WC & HF & GR & GDL & MI & LSC & BIO --> S[(241 skills)]
    SK --> S

Orchestrating Agents

Agents detect your need and route to the right skills. Each agent's page documents the skills it orchestrates and the workflow it runs.

AgentDomain
writing-assistantAny writing task — structure, revision, stickiness, pre-publish gate
scientific-writing-editorManuscripts, grants, letters, reviewer responses, career docs
superforecasterForecasting and probability via 5-phase calibrated pipeline
cognitive-design-architectCognitive design, information architecture, D3 viz, fallacy check
geometric-deep-learning-architectSymmetry discovery → group ID → equivariant architecture
math-intuition-coach3Blue1Brown-style geometric intuition for any ML/data math (attention, PCA, gradients, high-dim spaces)
literature-scan-coachSingle entry point for a weekly literature-scan project; routes weekly / catch-up / on-demand / re-synthesize intents and spawns paper-synthesizer per run with the right parameters
paper-synthesizerWorker for literature-scan-coach. One window, one digest. Fetches bioRxiv / medRxiv / PubMed / arXiv, filters, clusters by theme, writes the layered-reasoning synthesis with paper links
skill-creatorOrchestrates Adler's six-step methodology to transform a methodology document (PDF, markdown, paper, framework guide) into a properly structured Claude Code skill — SKILL.md + resources + rubric
graphrag-specialistKnowledge graph construction, embedding fusion, retrieval orchestration
company-analystEnd-to-end company valuation → buy/sell/hold recommendation
special-situations-analystDistressed / private / high-growth / financial-firm valuation
macroeconomic-analystDeep thematic research on one trend cluster — sub-trends, value chains, beneficiary archetypes, pricing-in assessment, watch-indicators
product-strategistOutside-in layered analysis of a real product — vision, strategy, tactics, operational surface, system & ML architecture map, strategist synthesis
capital-allocation-strategistFinancing mix, dividends/buybacks, project investment
acquisition-analystM&A standalone + synergy + max bid
ipo-strategistPre-IPO → post-IPO valuation and pricing range
mlb-fantasy-coachPrimary Yahoo Fantasy Baseball orchestrator; morning briefs
mlb-lineup-optimizerDaily start/sit with advocate + critic variants
mlb-waiver-analystWeekly add/drop and FAAB bid sizing
mlb-streaming-strategistWeekly pitching plan, two-start SPs, spot starts
mlb-trade-analyzerOn-demand trade offer evaluation with always-counter ladder
mlb-category-strategistWeekly push/punt plan across the 10 H2H categories
mlb-playoff-plannerJuly-onward positioning for weeks 21–23 playoff window
wc-directorFIFA World Cup Fantasy main-thread orchestrator — runs the evolutionary generation loop, presents decision boards, the manager decides
wc-strategistPopulation member — one archetype genotype per parallel spawn, builds a complete candidate squad/plan
wc-evolution-engineFitness, variance-clique selection, building-block crossover + repair, mutation, diversity guard
wc-synthesisLens fan-in reconciler — integrates advocate/critic verdicts, preserves residual dissent, never argmaxes
wc-scoutPlayer minutes/role/set-piece/fitness data spine via live web search
wc-fixture-analystFixture difficulty, progression odds (2026 best-thirds math), mismatches, swing calendar
wc-squad-architectFeasibility, value-per-million, transfers, Wildcard rebuild scoping, fragility scan
wc-matchday-tacticianXI, captain ladder, bench order by kickoff, manual-sub triggers — the in-round 20–40 pts edge
wc-chip-strategistTimes the five chips across the tournament horizon (deploy vs hold)
wc-ownership-analystEffective ownership, template vs differential, field over-concentration, mini-league rival leverage
household-cfoMaster orchestrator + synthesizer for the household finance team; runs per-drop, monthly, and ad-hoc chat
household-intake-classifierPer-PDF document classification + manifest + account matching
household-bookkeeperPDF → reconciled, deduplicated, categorized JSON store with append-only commit
household-spending-analystCategory trends, recurring detection, 60-day cash-flow forecast, seasonal calendar
household-bills-vigilanceMissed bills, duplicate / fraud / anomaly / fee scans with severity-tagged alerts
household-savings-debtGoals, emergency fund, debt strategy, mortgage prepay math, rewards optimization
household-investment-retirementPortfolio drift, contribution optimization, TLH with wash-sale awareness, retirement projection
household-tax-complianceDocument tracking, deduction tracker, HSA receipt vault, year-end packet for CPA
household-dashboard-designerWeekly static HTML dashboard with cognitive-design + storytelling + fallacy guard
librarianIngest + tag + index a Substack writer's corpus; topic ledger maintenance
intuition-builder5 distinct framings (everyday / physical / contrarian / historical / counterfactual) per technical topic
editorTwo-pass voice + structural review of drafts; voice gate before publish
distribution-translatorTurn published essays into Substack Note / X thread / LinkedIn post / cross-post blurb
growth-analystWeekly Substack CSV → rolling baseline → attribution → per-section report
trend-scoutWeekly ML/systems signal digest from ~35 curated intuition-first sources
technical-reviewerPre-publish claim check — simplified / wrong / contested / overclaim classification with primary sources
curatorEvery 4-6 weeks: section map maintenance; active + reactive; handles per-section voice overlays
growth-strategistQuarterly (rolling 13-week) strategic zoomout with uncomfortable questions and kill list
biostat-tutorSocratic experiential (Kolb) tutor for one learning module; never summarizes — draws the claim out and routes it to an evergreen note
biostat-assessorClosed-book Bloom-tagged probing; mastery banding; expanding-interval spaced-repetition scheduling
biostat-labLab + project mentor; designs and reviews experiential genomics projects; enforces cross-validation and predict-then-check
biostat-coachCurriculum coach; tracks phase/module progress and pacing; surfaces the spaced-repetition queue
biostat-scribeGenerative vault-style note/post writer from the learner's own evergreen claims (provenance-tracked)
biostat-editorAdvisory-only editor for dictated essays — directs structure, flags grammar, suggests language; never rewrites unless told (read-only by design)
biostat-emergenceBottom-up clustering of evergreen notes into structure notes (4+ to form a cluster, >12 to split)
biostat-healthVault + curriculum health audit — orphans, duplicate claims, weak titles, status drift, overdue reviews
biostat-vizInteractive D3 genetics/genomics visualizations for the docs/ site; delegates design review to cognitive-design-architect

Optional native dependencies

A small number of skills wrap native command-line tools and only work if those tools are installed on the user's machine. All such skills are optional. If a required tool is missing, the skill exits immediately with the exact install commands; nothing breaks silently and the rest of the library continues to work normally.

If you do not want to install the tools below, simply do not invoke the skills that need them. Every other skill in this repository, and every agent that does not include one of these skills in its pipeline, works without any additional native dependencies.

Skills with native dependencies

SkillNative dependencyWhy it's neededInstall (macOS)Install (Linux / Debian)
markdown-to-pdfpandoc and a LaTeX engine (xelatex)Renders a finished markdown report to an analyst-style PDF. The LaTeX engine produces the typography. Without these tools the skill cannot run; it has no fallback rendering path.brew install pandoc basictexsudo tlmgr update --selfsudo tlmgr install xetexsudo apt install pandoc texlive-xetex texlive-fonts-recommended

basictex on macOS is about 90 MB. Do not install mactex unless you specifically need the full LaTeX stack (~5 GB).

Agents that invoke these skills

Currently only one agent pipeline includes a dependency-requiring skill:

  • product-strategist calls markdown-to-pdf in its final step to render the finished strategist report.

If you run product-strategist without pandoc or xelatex installed, the agent still produces its primary markdown report; only the PDF rendering step is affected, and the agent will surface the install instructions in that case. You can install the tools later and re-render any past markdown report on demand.


Skills Index

241 skills across 8 super-categories. Every skill's full methodology, templates, and evaluation rubric live in its SKILL.md — click any entry to drill in.

Strategy & decision-making

Problem-solving & analysis

Estimation & forecasting

Dialogue & deliberation

Ideation & creativity

Productivity & learning

Research design & evaluation

Reading methodology (Adler-style; reusable across paper extraction, skill creation, evidence triage)

  • inspectional-reading — Level 1. First-level systematic skim — classify document type, decide whether deeper reading is worth it. Adler's "How to Read a Book" Level 1, generalized.
  • structural-analysis — Level 2. Classify content (practical vs theoretical, structure type, completeness 1-5), state unity in one sentence, enumerate parts, define problems. Adler's analytical-reading first stage.
  • component-extraction — Level 3. Section-by-section extraction of terms, propositions, arguments, solutions. Reading-strategy selection (section-based / windowing / targeted) for the right document size.
  • synthesis-application — Level 4. Completeness + logic + applicability gate after component extraction; produces GO / GO_WITH_GAPS / NO_GO verdict. Adler's "Is it true? What of it?" pass.
  • skill-construction — Construction step. Build SKILL.md + resources + rubric from extracted components at appropriate complexity (Level 1-3).
  • paper-three-pass-extraction — Three-pass + Five Cs extraction methodology specifically for academic papers. Wraps inspectional-reading as Pass 1; adds Pass 2 (content grasp via structural-analysis + component-extraction) and Pass 3 (deep understanding). Powers the paper-extractor agent.

Literature scan

Domain-neutral primitives for any weekly paper-digest workflow. Powers the literature-scan-coach + paper-extractor + paper-synthesizer pipeline.

  • fetch-preprint-recent — Fetch bioRxiv / medRxiv preprints for a date window with cursor pagination and client-side keyword filter.
  • fetch-pubmed-recent — Fetch PubMed records for a date window via PubMed MCP when available, E-utilities fallback otherwise.
  • fetch-arxiv-recent — Fetch arXiv papers for a date window with optional category restriction (cs.LG, cs.CL, stat.ML, q-bio.QM, etc.); Atom XML parsing.
  • paper-relevance-filter — Score candidate papers KEEP / REVIEW / DROP on match strength + criteria fit + novelty against last-4-weeks history.
  • paper-cluster-by-theme — Group filtered papers into 2-5 argument-shaped clusters before synthesis.

General writing

Scientific & academic writing

Architecture & product

  • adr-architecture — Document architecture decisions with context, options, consequences.
  • chain-spec-risk-metrics — Chain spec → risk → success metrics for launches.
  • roadmap-backcast — Plan backward from deadline to identify critical path and feasibility.
  • metrics-tree — Decompose a North Star metric into sub-metrics and leading indicators.

Design & information architecture

Security

Org & facilitation

Data modeling & visualization

Geometric deep learning

Math intuition

3Blue1Brown-style teaching moves for explaining vector / matrix / high-dim concepts geometrically. Powers the math-intuition-coach agent.

GraphRAG

Corporate finance & valuation (11)

Based on Damodaran's valuation curriculum.

Household finance (12)

PDF-driven personal-finance pipeline. Pairs with a companion orchestration runtime (inbox, archive, canonical JSON store, prompts, weekly static HTML dashboards). Powers the household-cfo agent and 8 specialists.

Game theory & strategic competition (8)

Domain-neutral primitives — portable across fantasy sports, poker, auctions, M&A.

Yahoo Fantasy Baseball (14)

Baseball-specific skills for a H2H Categories league. Pairs with a companion runtime workspace (league state, signals, decision trackers, daily briefs).

FIFA World Cup Fantasy (16)

Skills for the evolutionary World Cup fantasy backroom. Pairs with a companion runtime workspace (tournament state, signals, generations, decision boards).

  • wc-player-ev — Per-player per-round xEV: minutes cliff, fixture-scaled npxG/xA, pen/set-piece premium.
  • wc-clean-sheet-model — Clean-sheet probability + the stack-correlation bonus for defensive blocks.
  • wc-captain-ladder — Rolling-captaincy EV: optimal armband switching across staggered kickoffs.
  • wc-fixture-progression — Two-way fixture difficulty, progression odds (incl. 2026 best-thirds), mismatches, swings.
  • wc-ownership-meta — Effective ownership, template vs differential, rank-leverage under protect/gain.
  • wc-matchday-timing — Kickoff spread, bench order by kickoff, manual-sub triggers, MD3 simultaneity guard.
  • wc-transfer-planner — Free-transfer ROI, forced elimination swaps, Wildcard-rebuild threshold.
  • wc-chip-timing — Per-chip leverage by round; earmarks, triggers, the never-hoard-to-the-semis clock.
  • wc-fitness-eval — Risk-adjusted, rank-relative fitness with full decomposition + Goodhart guard.
  • wc-building-block-crossover — Recombine squads at building-block boundaries with a feasibility repair operator.
  • wc-archetype-mutation — Bounded, feasibility-preserving perturbations on offspring candidates.
  • wc-population-diversity — Anti-inbreeding guard: collapse metrics, wider-kernel re-seed, selection-pressure tuning.
  • wc-decision-board — Render the advisory decision board: 2–4 options, decomposition, dissent, your-call.
  • wc-tournament-state — The tournament state machine: phase, budget, nation cap, chips, elimination horizons.
  • wc-decision-logger — Log the option set + pick + reasoning; archive generations; update scoreboards.
  • wc-signal-emitter — Validate and persist signals with schema, range, and citation enforcement.

Specialized & meta (2)

  • chef-assistant — Cook via technique, food science, flavor architecture.
  • skill-creator — Turn a PDF / paper / methodology into a reusable skill.

Pairs with a companion working folder (inbox, corpus, shared-context, ops). Voice-profile, section-map, per-section visual identities, analogy catalog, topic ledger all live there.

Librarian (8) — corpus ingest + index

  • ingest-inbox-item — Orchestrate per-file ingest: normalize → tag → score → dedupe → write seed.
  • normalize-format — Convert markdown / Claude exports / transcripts / Readwise into canonical seed bodies.
  • tag-by-topic — Assign 1-4 topic tags from controlled vocabulary; log pending tags.
  • score-intuition-density — Compute 0-10 density from 8 explicit signals (analogy, concrete example, counterfactual, etc.).
  • dedupe-against-corpus — Exact fingerprint + near-match Jaccard; link rather than merge.
  • search-corpus — "What have I already thought about X?" — ranked matches with excerpts.
  • update-topic-ledger — Maintain counts, last_touched, hot/warm/cold, top_seeds per topic.
  • sweep-stale-seeds — Flag seeds >30d old with no links; recommend keep/promote/kill.

Intuition Builder (7) — 5 framings per topic

Editor (10) — voice + structural review gate

Distribution Translator (7) — platform-native rewrites

Growth Analyst (8) — weekly Substack stats

Trend Scout (6) — weekly ML/systems signal digest

Technical Reviewer (6) — pre-publish claim check

Curator (9) — section-map maintainer, active + reactive

Growth Strategist (7) — quarterly zoomout

Pairs with a companion learning vault (Kolb curriculum, Zettelkasten evergreen notes, dictation-to-publish writing studio, GitHub Pages site). Powers the 9 biostat-* agents.

  • experiential-kolb-teaching — Run a learning module as a Kolb cycle; withhold the explanation, draw the claim out, route it to an evergreen note.
  • mastery-assessment — Bloom-tagged closed-book probing + Dreyfus mastery bands + expanding-interval spaced repetition.
  • zettel-note — Write atomic declarative-claim vault notes with piped links and search-before-create dedup.
  • learning-in-public-voice — House style for learning-in-public essays; defers to the writer's own voice-profile.md.
  • advisory-edit — Strict advisory-only editing: direct structure, flag grammar, suggest language; never rewrite unless told.
  • spoken-draft-cleanup — Turn a dictation transcript into a clean draft with the voice preserved; mechanical-only, flag don't fix.
  • genomics-viz — Build interactive D3 genetics visualizations (HWE, drift, LD, GxE, prediction accuracy) with visual-storytelling.

Installation

Option 1 — Plugin (recommended)

/plugin marketplace add lyndonkl/claude
/plugin install thinking-frameworks-skills

All 241 skills become available immediately. Claude invokes them automatically based on your request and each skill's trigger description.

Clone and copy into your Claude Code skills directory:

git clone https://github.com/lyndonkl/claude.git
cd claude

# macOS / Linux
cp -r skills/* ~/.claude/skills/

# Windows
xcopy skills\* %USERPROFILE%\.claude\skills\ /E /I

Using Skills

Skills activate automatically when you use matching trigger phrases — no manual invocation needed.

"Help me decide between three options..."      → decision-matrix
"I need to run user interviews..."              → discovery-interviews-surveys
"Value this company for acquisition..."         → acquisition-analyst (agent)
"What should I stream at pitcher this week?"    → mlb-streaming-strategist (agent)
"My chicken keeps coming out dry..."            → chef-assistant
skill-name/
├── SKILL.md                          # Entry point (< 300 lines)
│   ├── YAML metadata with triggers
│   ├── When to use + guardrails
│   ├── Workflow checklist
│   └── Common patterns
└── resources/
    ├── template.md                   # Ready-to-use templates (< 400 lines)
    ├── methodology.md                # Advanced techniques (< 500 lines)
    └── evaluators/
        └── rubric_[skill].json       # 10-criterion quality rubric

Progressive depth: SKILL.md for the skim, template.md for the copy-paste, methodology.md for the edge cases, rubric for self-evaluation. Every skill ships with all four.


Contributing

Fork, branch, and PR. Follow the skill structure above — SKILL.md under 300 lines, self-assess against the rubric (target ≥ 3.5), and include trigger phrases in the YAML description so Claude can auto-invoke.

Resources

License

Open source; use freely with Claude Code. Individual skills carry their own attributions.

Acknowledgments

Skills draw on established frameworks and practitioners:

  • Thinking: Bayesian reasoning, causal inference, dialectics, first principles
  • Research: Jobs-to-be-done (Christensen), design of experiments (Box, Taguchi), user research (Torres, Ulwick)
  • Decision-making: Multi-criteria analysis, expected value, risk analysis
  • Corporate finance: Damodaran's valuation curriculum
  • Writing: Story structure (McKee, McPhee), clarity (Zinsser, Pinker, King), SUCCESs (Heath)
  • Culinary: Technique (Pépin, Child, López-Alt), food science (McGee, Sharma), cultural cooking (Bourdain, Chang, Ottolenghi)
Ver o original no GitHub

Bom para

Relacionados