Skip to content
BranchesTags

Latest commit

 

History

205 Commits

Folders and files

NameName
Last commit message
Last commit date
.claude-plugin
.claude-plugin
 
 
agents
agents
 
 
hooks
hooks
 
 
knowledge
knowledge
 
 
skills
skills
 
 
.gitignore
.gitignore
 
 
README.md
README.md
 
 

Repository files navigation

Claude Code Skills Collection

Skills Agents Status Run in Smithery

A production-ready library of 271 skills and 85 orchestrating agents for Claude Code.

It covers thinking frameworks, research, writing, design, and data/ML. It covers game theory and household personal finance. Larger teams handle specific domains: a 15-agent corporate finance and valuation suite, a 10-agent FIFA World Cup fantasy backroom, a 9-agent team for growing a Substack publication, and a 9-agent studio for learning ML-driven crop genetics. Also included: five reusable primitives for gated, verifiable research pipelines, and a 3-agent narrative team that turns researched evidence into long-form nonfiction.

Install in 30 seconds:

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

Note: Two skills need something installed separately: markdown-to-pdf needs pandoc and a LaTeX engine, and readability-check needs the textstat Python package. See Optional dependencies below. Every other skill works 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 IPO / evaluate a project valuation-orchestrator — one entry point, routes to 14 specialists
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 from public material, explained so a newcomer can follow it product-strategist
Decide how to allocate capital (debt / dividends / projects) valuation-orchestrator in corporate-finance mode
Manage my Yahoo Fantasy Baseball team mlb-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 letter scientific-writing-editor
Improve any piece of writing (blog, memo, essay) writing-assistant
Turn a pile of research into structured long-form nonfiction (find the shape, then write it) narrative-architect → scenewright, with writing-assistant
Make a calibrated forecast or probability estimate superforecaster
Design a dashboard, viz, or UI grounded in cognition cognitive-design-architect
Build a GraphRAG / knowledge-graph retrieval system graphrag-specialist
Design equivariant / symmetry-aware neural networks geometric-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 watchlist literature-scan-coach → paper-synthesizer
Grow my Substack / publish intuition-first ML & systems essays librarian, 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
Build a personalized schedule for a conference (ingest the program → cluster themes → a few grounded questions → optimized plan) conf-director + 5 specialists
Run a gated research pipeline where every number is sourced, attacked, and verified before anything is built on it market-era-historian, series-archaeologist, mechanism-analyst, claim-verifier, stage-auditor
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 -->|Value / analyze a company| CA[valuation-orchestrator<br/>+ 14 finance specialists]
    D -->|Analyze a product| PS[product-strategist]
    D -->|Write something| WA[writing-assistant<br/>scientific-writing-editor]
    D -->|Structure + write research| NA[narrative-architect<br/>scenewright]
    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 -->|Plan a conference| CONF[conf-director<br/>+ 5 specialists]
    D -->|Gated research pipeline| RP[market-era-historian<br/>series-archaeologist<br/>mechanism-analyst<br/>claim-verifier<br/>stage-auditor]
    D -->|One-off tool| SK[Skills Index ▾]
    CA & PS & WA & NA & SF & CD & MLB & WC & HF & GR & GDL & MI & LSC & BIO & CONF & RP --> S[(262 skills)]
    SK --> S
Loading

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.

Agent Domain
writing-assistant Any writing task — structure, revision, stickiness, pre-publish gate
scientific-writing-editor Manuscripts, grants, letters, reviewer responses, career docs
narrative-architect Researched corpus → validated narrative architecture. Evidence ledger, form triage, designing principle, opposition web, beat map. Selects the structure (or reports that the material has none) and never drafts
scenewright Executes an architecture into prose. SCAM scenes, POV and distance, ladder of abstraction, number handling, Hart's nine line passes. Source-bound, and escalates what it cannot fix at its own layer
superforecaster Forecasting and probability via 5-phase calibrated pipeline
cognitive-design-architect Cognitive design, information architecture, D3 viz, fallacy check
geometric-deep-learning-architect Symmetry discovery → group ID → equivariant architecture
math-intuition-coach 3Blue1Brown-style geometric intuition for any ML/data math (attention, PCA, gradients, high-dim spaces)
literature-scan-coach Single 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-synthesizer Worker 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-creator Orchestrates 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-specialist Knowledge graph construction, embedding fusion, retrieval orchestration
valuation-orchestrator Entry point for all company analysis. Classifies the company, enforces the gates, routes to 14 specialists, reconciles the verdict
company-diagnostician Classifies the company and compiles the hard constraints that bind every later stage
financial-data-collector Gathers filings, market data and reference tables; reports gaps honestly
business-narrative-analyst Builds the business story and converts it into valuation drivers
financial-statement-analyst Capitalizes leases and R&D, strips one-offs, restates EBIT and capital together
cost-of-capital-analyst Riskfree rate, equity risk premium, bottom-up beta, synthetic rating, WACC
intrinsic-valuation-analyst Runs the DCF to a value per share, then says what the market price assumes
relative-valuation-analyst Prices against peers and the market; says when a multiple is not usable
special-situations-analyst Banks, pre-revenue, distressed, private, cyclical — the branches standard DCF cannot handle
capital-structure-analyst How much debt the firm should carry, and what kind
payout-policy-analyst Whether the company returns the right amount of cash to its owners
investment-analyst Projects and acquisitions on incremental cash flows; control and synergy
real-options-analyst Values genuine optionality, and rejects the candidates that only look like options
valuation-critic Attacks the finished analysis and raises findings; never edits the work it reviews
investment-reconciler Reconciles every strand into a verdict and writes the report
macroeconomic-analyst Deep thematic research on one trend cluster — sub-trends, value chains, beneficiary archetypes, pricing-in assessment, watch-indicators
product-strategist Outside-in analysis of a real product — vision, strategy, tactics, operational surface, strategist verdict. Every bet graded by the evidence behind it, every term unpacked for a reader with no background
mlb-fantasy-coach Primary Yahoo Fantasy Baseball orchestrator; morning briefs
mlb-lineup-optimizer Daily start/sit with advocate + critic variants
mlb-waiver-analyst Weekly add/drop and FAAB bid sizing
mlb-streaming-strategist Weekly pitching plan, two-start SPs, spot starts
mlb-trade-analyzer On-demand trade offer evaluation with always-counter ladder
mlb-category-strategist Weekly push/punt plan across the 10 H2H categories
mlb-playoff-planner July-onward positioning for weeks 21–23 playoff window
wc-director FIFA World Cup Fantasy main-thread orchestrator — runs the evolutionary generation loop, presents decision boards, the manager decides
wc-strategist Population member — one archetype genotype per parallel spawn, builds a complete candidate squad/plan
wc-evolution-engine Fitness, variance-clique selection, building-block crossover + repair, mutation, diversity guard
wc-synthesis Lens fan-in reconciler — integrates advocate/critic verdicts, preserves residual dissent, never argmaxes
wc-scout Player minutes/role/set-piece/fitness data spine via live web search
wc-fixture-analyst Fixture difficulty, progression odds (2026 best-thirds math), mismatches, swing calendar
wc-squad-architect Feasibility, value-per-million, transfers, Wildcard rebuild scoping, fragility scan
wc-matchday-tactician XI, captain ladder, bench order by kickoff, manual-sub triggers — the in-round 20–40 pts edge
wc-chip-strategist Times the five chips across the tournament horizon (deploy vs hold)
wc-ownership-analyst Effective ownership, template vs differential, field over-concentration, mini-league rival leverage
household-cfo Master orchestrator + synthesizer for the household finance team; runs per-drop, monthly, and ad-hoc chat
household-intake-classifier Per-PDF document classification + manifest + account matching
household-bookkeeper PDF → reconciled, deduplicated, categorized JSON store with append-only commit
household-spending-analyst Category trends, recurring detection, 60-day cash-flow forecast, seasonal calendar
household-bills-vigilance Missed bills, duplicate / fraud / anomaly / fee scans with severity-tagged alerts
household-savings-debt Goals, emergency fund, debt strategy, mortgage prepay math, rewards optimization
household-investment-retirement Portfolio drift, contribution optimization, TLH with wash-sale awareness, retirement projection
household-tax-compliance Document tracking, deduction tracker, HSA receipt vault, year-end packet for CPA
household-dashboard-designer Weekly static HTML dashboard with cognitive-design + storytelling + fallacy guard
librarian Ingest + tag + index a Substack writer's corpus; topic ledger maintenance
intuition-builder 5 distinct framings (everyday / physical / contrarian / historical / counterfactual) per technical topic
editor Two-pass voice + structural review of drafts; voice gate before publish
distribution-translator Turn published essays into Substack Note / X thread / LinkedIn post / cross-post blurb
growth-analyst Weekly Substack CSV → rolling baseline → attribution → per-section report
trend-scout Weekly ML/systems signal digest from ~35 curated intuition-first sources
technical-reviewer Pre-publish claim check — simplified / wrong / contested / overclaim classification with primary sources
curator Every 4-6 weeks: section map maintenance; active + reactive; handles per-section voice overlays
growth-strategist Quarterly (rolling 13-week) strategic zoomout with uncomfortable questions and kill list
biostat-tutor Socratic experiential (Kolb) tutor for one learning module; never summarizes — draws the claim out and routes it to an evergreen note
biostat-assessor Closed-book Bloom-tagged probing; mastery banding; expanding-interval spaced-repetition scheduling
biostat-lab Lab + project mentor; designs and reviews experiential genomics projects; enforces cross-validation and predict-then-check
biostat-coach Curriculum coach; tracks phase/module progress and pacing; surfaces the spaced-repetition queue
biostat-scribe Generative vault-style note/post writer from the learner's own evergreen claims (provenance-tracked)
biostat-editor Advisory-only editor for dictated essays — directs structure, flags grammar, suggests language; never rewrites unless told (read-only by design)
biostat-emergence Bottom-up clustering of evergreen notes into structure notes (4+ to form a cluster, >12 to split)
biostat-health Vault + curriculum health audit — orphans, duplicate claims, weak titles, status drift, overdue reviews
biostat-viz Interactive D3 genetics/genomics visualizations for the docs/ site; delegates design review to cognitive-design-architect
conf-director Conference-schedule pipeline orchestrator — runs ingest → cluster → elicit → schedule, gates on per-stage confidence, guards its invariants, surfaces conflicts, never auto-commits (reusable across conferences)
conf-program-ingestor Stage 1 — parses a conference program into confidence-scored event records + axes; fans out enrichment
conf-enrichment-researcher Isolated web-research sub-worker — thickens one thin abstract with provenance + confidence
conf-theme-cartographer Stage 2 — embed-then-label clustering into 6–8 themes + outlier bucket + soft multi-membership affinities
conf-preference-elicitor Stage 3 (interactive) — a few grounded, choice-based questions over an uncertainty region, with deliberate outlier probes; builds the preference profile
conf-schedule-optimizer Stage 4 — constraint optimization under user-owned weights; protects contiguous free time; surfaces unbreakable conflicts
market-era-historian Reconstructs how a market worked in one bounded period — fills a provided field schema with sourced, calibrated claims; built for parallel fan-out across periods
series-archaeologist Assembles decades-long datasets from incompatible sources — per-point provenance, a concordance of definitional breaks, documented bridges, no silent splices
mechanism-analyst Worked-number analysis of a market mechanism or deal economics vs named rivals — machine-checkable arithmetic plus a simulation-ready parameterization
claim-verifier Refute-first adversarial verification of a claim batch — independent re-sourcing, re-derivation, interval and grade challenge; confirmed / adjusted / rejected / unverified
stage-auditor The automated gate between pipeline stages — runs a contract's deterministic checks, judges the rest, emits pass/fail plus a remediation payload; audits, never repairs

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

Skill Native dependency Why it's needed Install (macOS) Install (Linux / Debian)
markdown-to-pdf pandoc 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 basictex
sudo tlmgr update --self
sudo tlmgr install xetex		 sudo apt install pandoc texlive-xetex texlive-fonts-recommended
readability-check textstat (Python package) Computes the five readability formulas. Without it the skill prints the exact install command and exits; it does not guess scores. python3 -m pip install --user textstat python3 -m pip install --user textstat

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

271 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.

🧠 Thinking & Decisions — decision-making, problem-solving, estimation, dialogue, ideation, learning (37 skills)

Strategy & decision-making

Problem-solving & analysis

Estimation & forecasting

Dialogue & deliberation

Ideation & creativity

Productivity & learning

🔬 Research & Evidence — research design, evidence evaluation, rubrics, ethics, reading methodology, literature scan (17 skills)

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.
✍️ Communication & Writing — writing pipeline, narrative architecture and craft, scientific writing, audience adaptation, analyst voice, PDF rendering (29 skills)

General writing

Narrative architecture & craft (14)

The toolkit behind narrative-architect and scenewright. Built on Truby (structure), Hart (scene and line craft), and McPhee (arrangement), and hardened against the failure they share — a fluent model filling every empty slot whether the evidence supports it or not. Works when the protagonist is a person and when it is a system: a market, a protocol, an institution, a supply chain.

Macro — deciding the shape:

  • narrative-evidence-ledger — Build the ledger before any structure is chosen. Frame lock, evidence classes, causal triage L1–L5, one-way finding→claim→beat promotion, and a DEAD column that keeps what the story is not.
  • narrative-form-triage — The refusal gate. Six questions and a thirteen-structure table decide story, explanatory, gathering, or do not narrate. "This material has no story" is a passing verdict.
  • narrative-arc-mapping — Truby's steps mapped onto evidence, with step-fit triage and the empty-slot protocol. Every slot filled is a defect, not a triumph.
  • narrative-opposition-web — Four-corner opposition, the best-possible-opponent audit, and the warrant rule that stops whoever lost being cast as the villain.
  • mcphee-structure-derivation — McPhee's coding-and-sorting pass: code every chunk, sort by chronology and by theme, resolve the disagreements, name the shape. Derives a structure when no dramatic engine exists; hands off to writing-structure-planner.
  • systemic-protagonist — Render a market, protocol, or institution as a protagonist without faking agency. POSIWID character sheet, the agency ledger's four verb repairs, gateway proxies, climax by constraint violation.

Micro — executing it:

  • scene-construction-scam — SCAM scene qualification with a provenance gate, scene floor and ceiling, the 1-of-20 detail cut. An unfillable slot demotes to summary rather than being invented.
  • ladder-of-abstraction — Rung tagging R1–R4, the middle-rung trap, and the honest ceiling. Routes a readability failure to a concrete example rather than to synonym substitution.
  • prose-force-and-rhythm — Hart's nine Wordcraft passes in strict order with upward escalation, the claim-strength invariant, and a protected-hedge list so brevity never deletes the uncertainty.
  • reader-first-prose — Rewrites for a reader meeting the subject for the first time: introduce every term before use, ground or cut each metaphor, plain over clever, no em dash and no terse "X. Not Y." antithesis. Keeps inference marked as inference so the prose never out-claims its source.
  • term-interrogation — Walks a finished document line by line, finds every word standing in for a concept, and unpacks each one recursively until it bottoms out in plain language. Targets the assumed-knowledge phrase ("moat", "flywheel", "at scale") that sounds ordinary while hiding the argument. Emits a term ledger and a plain-language glossary.
  • numbers-in-narrative — Land a quantity in a sentence without overstating it. Denominator lock, uncertainty routing, the likelihood/confidence lexicon, chart-beat contracts.

Gates — used by both:

  • narrative-fidelity-audit — The bright lines an agent must refuse to cross, the interior-state provenance ladder, an LLM-specific fabrication scan, and the "how do you know?" pass at sentence, paragraph and passage level.
  • narrative-fallacy-guard — Stops a true set of facts becoming a false story. Retrospective slot audit, outcome-blind rewrite, inevitability audit, survivorship, proportion and omission. Labels and discloses; never deletes.
  • narrative-handoff-contract — The machine-readable contract between the two agents, plus a validator that enforces it. Evidence arrays over values, ABSENT as a passing state, and an escalation object for sending defects back upward.

Scientific & academic writing

🏗️ Building & Shipping — architecture, design, security, org, product delivery (15 skills)

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 & ML — data modeling, visualization, math intuition, GraphRAG, geometric deep learning (19 skills)

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

💼 Domain Packs — corporate finance, household finance, game theory, fantasy baseball, conference scheduling, specialized (53 skills)

Corporate finance & valuation (19)

Distilled from Aswath Damodaran's corporate finance and valuation curriculum. The source material and the design record live in knowledge/: 6 playbooks, 362 concept notes, 23 reverse-engineered spreadsheet models, and notes on all 2,424 lecture pages. The ten computation skills ship tested Python that runs on a bare interpreter — no third-party packages — so the agents choose and defend inputs while scripts do the arithmetic.

Computation — each ships a script with a selftest subcommand:

Method — knowledge and procedure, no scripts:

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.

Conference scheduling (6)

Reusable, conference-agnostic pipeline for building a personalized conference schedule. Powers the conf-director agent + 5 specialists; pairs with a per-conference workflow folder (config, data-flow handoff dirs, state machine, runbook). Built around explicit per-stage uncertainty: confidence travels with the data and the system surfaces the decisions only the attendee can make.

  • conf-program-extraction — Parse any conference program into normalized event records with per-field confidence and independent axes (topic, depth, format, prereqs, recorded, capacity).
  • conf-abstract-enrichment — Thicken thin / missing abstracts via targeted web search, gated, with provenance and confidence on every recovered claim.
  • conf-theme-clustering — Embed-then-label clustering into 6–8 coarse themes + an outlier bucket + soft multi-membership affinities (+ cluster.py, with an LLM-reasoned fallback).
  • conf-preference-elicitation — A few grounded, choice-based questions over a preference region; information-gain question selection, explore/exploit, and mandatory outlier probes against selection bias.
  • conf-schedule-optimization — Constraint optimization under user-owned weights (interest / breadth / pacing / serendipity); protects contiguous free time and surfaces unbreakable conflicts (+ schedule.py).
  • conf-pipeline-orchestration — The staged-pipeline orchestration discipline: confidence propagation, structured artifact contracts, invariant-guard, the diversity floor, concentrated hardening, and the Goodhart caution.

Specialized & meta (2)

  • chef-assistant — Cook via technique, food science, flavor architecture.
  • skill-creator — Turn a PDF / paper / methodology into a reusable skill.
📰 Substack Growth — 9-agent team for growing "From First Principles / The Thinker's Notebook" (68 skills)

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

📚 Learning Studio — experiential study + learning-in-public writing for ML-driven crop genetics (7 skills)

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

Option 2 — Manual install

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..."         → valuation-orchestrator (agent)
"What should I stream at pitcher this week?"    → mlb-streaming-strategist (agent)
"My chicken keeps coming out dry..."            → chef-assistant
Skill anatomy — how each skill is structured 
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)

About

Agents, skills and anything else to use with claude

Resources

Readme
Activity

Stars

149 stars

Watchers

0 watching

Forks

23 forks
Report repository

Releases

Packages

Contributors

Languages