Reference · your Claude Code toolkit
69 skills in ~/.claude/skills/ · generated 2026-06-20 · printable A4 reference
How a skill fires: 🔵 you type the slash command · 🟢 Claude auto-invokes when your words match the description · 🟡 a hook runs it automatically · 🟣 it only runs when another skill calls it or you invoke it explicitly. Skills cannot call other skills programmatically — “Related” means Claude will suggest the next one, or it's invoked as a subagent.
plans/<slug>/ + a copy-paste prompt.| Build a non-trivial feature across many files | plan-explore → plan-techspec → execute |
| Tiny one-file fix | just do it — skip the chain |
| Understand how existing code is wired | graphify (query first) |
| Stress-test a plan before building | grilling / grill-me |
| Check how a page looks / matches a mockup | visual-qa (+ polish-pass first) |
| Prove a whole project surface works | qa-suite |
| Review a branch before merge | review → ultra-review for risky cuts |
| A hard bug or perf regression | diagnosing-bugs |
| Start something brand new | new-project |
| Close out a session | finish-session (or -tts for the audio course) |
| Issue a Hungarian invoice | szamlazz-invoice |
| Draft a client contract | create-contract → docuseal-cli |
| Find local-business leads | gmaps-leads → instantly-campaigns |
| Turn notes into an article | writing-shape / writing-beats |
| Make narration / a voiceover | tts |
| Which model/effort to use | model-help |
| Where a credential / env var lives | house-reference |
| Make an AI output measurably reliable | autoresearch-eval-lab + autoresearch-loops |
Turn a fuzzy idea into an executed, test-verified change across 3 fresh chats. Each instance hands off to the next via files in plans/<slug>/ + a copy-paste prompt.
| Skill | Trigger | What it does | Inputs → Outputs | Related |
|---|---|---|---|---|
| 🔵 plan-explore | /plan-explore · “plan this” · “scope this change” | Instance 1. Explores the codebase (graphify-first), interviews you until intent is locked, writes a self-contained plan bundle. | fuzzy intent + repo → HANDOFF.md · manifest.json · PROMPT_to_techspec.md | graphify grilling user-journeys plan-techspec |
| 🔵 user-journeys | /user-journeys · “journey map” · “behavioral spec” | Optional Instance-1 companion. Click-by-click behavioral spec of every user path → an HTML doc + a test-ID bridge. | HANDOFF.md / chat → journeys.html (+ --offline) · scenario_map.json | plan-explore cloudflare-staging plan-techspec |
| 🔵 plan-techspec | /plan-techspec · “tech spec this” · “research + checklist” | Instance 2. Researches the approach (subagents), writes the excruciatingly-detailed red→green→refactor checklist.md + ADRs, then calls build-qa-skill. | HANDOFF.md (+ scenario_map.json) → checklist.md · ADRs · execute_context.json · PROMPT_to_execute.md | graphify build-qa-skill grilling |
| 🟣 build-qa-skill | Called by plan-techspec (or /build-qa-skill) | META skill. Copies the qa-harness engine into the target repo and wires one check per checklist ID; runs a post-write --list smoke. | checklist.md → a project-local QA skill in <repo>/.claude/skills/ | qa-harness skill-creator qa-suite plan-techspec |
| 🟣 qa-harness | Copied by build-qa-skill (rarely run directly) | TEMPLATE engine — two-file config-driven Playwright harness (capture.py safety + runner.py checks). Sentinel-gated + send-blocklisted. | config.json + fixtures → PASS/FAIL/REFUSE/STUB per check-ID | build-qa-skill qa-suite |
| 🔵 qa-suite | /qa-suite · “run the QA suite” · “full test pass” | Instance-3 whole-surface verification. Detects the stack, runs browser/visual/api/backend layers, rolls up GREEN/RED. | repo → layer roll-up table (OVERALL GREEN|RED) | qa-harness visual-qa |
| 🟢 grilling | “grill me” · “stress-test this plan” | Relentless one-question-at-a-time interview to sharpen a plan/design before building. | a plan/design → hardened intent | plan-explore grill-me grill-with-docs |
| 🟣 grill-me | explicit invoke | Lean relentless interview variant (no docs side-effects). | a plan → sharpened plan | grilling grill-with-docs |
| 🟣 grill-with-docs | explicit invoke | Relentless interview that also writes ADRs + a glossary as it goes. | a plan → sharpened plan + ADRs + glossary | grilling domain-modeling |
| 🟣 handoff | explicit invoke | Compacts the current conversation into a cold-start handoff doc for another agent. | current chat → handoff.md | plan-explore |
| 🟣 decision-mapping | explicit invoke | Turns a loose idea into a sequenced map of investigation tickets, then drives them one at a time. | loose idea → ordered investigation tickets → resolutions | to-issues triage |
| 🟢 request-refactor-plan | “plan a refactor” · “refactoring RFC” | Interview → a detailed refactor plan of tiny commits, filed as a GitHub issue. | refactor intent → GitHub issue (incremental steps) | review to-issues |
| 🟣 to-prd | explicit invoke | Synthesizes the current conversation into a PRD on the issue tracker (no interview). | chat → PRD | to-issues implement |
| 🟣 to-issues | explicit invoke | Breaks a plan/PRD into independently-grabbable issues as tracer-bullet vertical slices. | plan/PRD → issues | to-prd triage implement |
| 🟣 triage | explicit invoke | Moves issues/PRs through a triage state machine — categorise, verify, grill, write agent-ready briefs. | raw issues/PRs → triaged briefs | to-issues review |
| 🟣 implement | explicit invoke | Implements a piece of work from a PRD or set of issues. | PRD/issues → code | to-prd to-issues tdd |
Shape interfaces and domain language before (or instead of) writing code.
| Skill | Trigger | What it does | Inputs → Outputs | Related |
|---|---|---|---|---|
| 🟢 codebase-design | “design this module” · “where does the seam go” | Shared vocabulary for deep modules — find deepening opportunities, place seams, make code testable/AI-navigable. | a module → interface design notes | domain-modeling improve-codebase-architecture |
| 🟢 domain-modeling | “pin down the domain” · “ubiquitous language” | Build and sharpen a project's domain model + record architectural decisions. | conversation → domain model + ADRs | ubiquitous-language grill-with-docs |
| 🟣 ubiquitous-language | explicit invoke | Extracts a DDD glossary from the conversation, flags ambiguities, proposes canonical terms. | chat → UBIQUITOUS_LANGUAGE.md | domain-modeling |
| 🟢 design-an-interface | “design it twice” · “explore interface options” | Generates multiple radically different interface designs for a module via parallel sub-agents. | a module → N candidate interfaces | codebase-design prototype |
| 🟣 prototype | explicit invoke | Builds a throwaway prototype — a runnable terminal app for logic, or several toggleable UI variations. | a design question → runnable throwaway | design-an-interface |
| 🟣 improve-codebase-architecture | explicit invoke | Scans a codebase for deepening opportunities, shows a visual HTML report, then grills the one you pick. | codebase → HTML report → grilled plan | codebase-design grilling |
Prove the change works and is clean before it ships.
| Skill | Trigger | What it does | Inputs → Outputs | Related |
|---|---|---|---|---|
| 🟢 tdd | “test-first” · “red-green-refactor” | Test-driven development loop — write the failing test, make it pass, refactor. | a feature/bug → tests + passing code | qa-suite implement |
| 🟢 qa | “QA session” · “report bugs” · “file issues” | Conversational QA — you report bugs, it explores the code for context and files GitHub issues. | bug reports → GitHub issues | triage diagnosing-bugs |
| 🟢 visual-qa | “how does this page look” · “compare to the mockup” | CSV-driven visual QA — screenshots a page section-by-section at 3 viewports, scores 24 checks, auto-writes a fix prompt, iterates ≥3 rounds. | a URL/build → scored CSV + fix prompts → clean page | polish-pass qa-suite chrome-devtools |
| 🟢 polish-pass | “polish this” · “anti-AI-slop audit” | Final finishing pass on a built frontend — catches the tells that read as “AI made that” and fixes them. Run before visual-qa. | a built frontend → de-slopped frontend | visual-qa |
| 🟢 review | “review this branch/PR” | Reviews changes on two axes — Standards (repo coding standards) + Spec (matches the issue/PRD) — in parallel sub-agents. | a diff/branch → side-by-side review | ultra-review triage |
| 🔵 ultra-review | /ultra-review (alias for /code-review ultra) | Deep multi-agent cloud review of the branch diff — correctness, security, performance, readability. Heavier than review. | branch diff → deep cloud review | review |
| 🟢 diagnosing-bugs | “diagnose” · “debug this” · “it's failing/slow” | Structured diagnosis loop for hard bugs + performance regressions. | a broken/slow symptom → root cause + fix | qa graphify |
Drive the browser, ship static sites, and push code to the cloud.
| Skill | Trigger | What it does | Inputs → Outputs | Related |
|---|---|---|---|---|
| 🟢 chrome-devtools | any browser automation / screenshots / scraping | Operating playbook for the Chrome DevTools MCP — the ONLY sanctioned browser tool (never Puppeteer). | a web task → driven browser | visual-qa cloudflare-staging |
| 🟢 cloudflare-staging | “stage/deploy/share a preview link” | Deploys a static build folder to Cloudflare Pages staging, returns a client-ready *.pages.dev URL. | a build dir → *.pages.dev URL | user-journeys send-to-dani visual-qa |
| 🟢 local-server | “run locally” · “test webhooks locally” | Runs the Claude orchestrator locally with Cloudflare tunnelling. | local code → tunnelled live URL | modal-deploy add-webhook |
| 🟢 modal-deploy | “deploy to Modal” · “push to cloud” | Deploys execution scripts / functions to Modal cloud. | app code → live Modal deployment | add-webhook local-server |
| 🟢 add-webhook | “create a webhook” · “add an endpoint” | Adds new Modal webhooks for event-driven execution. | an event spec → a live webhook endpoint | modal-deploy |
Open and close sessions cleanly; keep repos to the house standard.
| Skill | Trigger | What it does | Inputs → Outputs | Related |
|---|---|---|---|---|
| 🟡 start-session | auto (SessionStart hook) · “start the session” | Source-gated git sync, loads PROJECT_STATE + newest CHANGELOG + GOALS + the registry card, prints a “you are here” briefing. | a repo → session briefing | finish-session document-changelog |
| 🔵 finish-session | /finish-session · “wrap up” · “commit and document” | Disciplined close-out — tidy + secret-scan, document-changelog, maintain-repo, curate ideas, ONE branch-safe commit, End-of-Execution report. | a changed repo → tidy + docs + commit + report | document-changelog maintain-repo finish-session-tts |
| 🔵 finish-session-tts | /finish-session-tts | Heavier finish-session — also refreshes the chat-history graph and produces a ~40-min personal audio course + HTML digest to Telegram. | a session → commit + docs + audio course | finish-session set-up-claude-chat-history-graph tts |
| 🟢 document-changelog | “document the session” · “update the changelog” | Prepends a curated CHANGELOG.md entry + refreshes PROJECT_STATE.md. Used by finish-session. | changes → CHANGELOG + PROJECT_STATE | finish-session maintain-repo |
| 🟢 maintain-repo | “tidy the repo” · “refresh the maps” · “check drift” | Audits a repo's information architecture + regenerates its living HTML maps. Needs the .repo-maps standard. | a .repo-maps repo → fresh maps + drift audit | restructure-repo new-project |
| 🟢 restructure-repo | “restructure/reorganize a repo” | Brings an unconfigured repo up to the .repo-maps standard with minimal change + zero breakage (additive retrofit). | a messy repo → .repo-maps-standard repo | maintain-repo new-project |
| 🟣 new-project | explicit invoke | The one skill for starting anything new — scaffolds the Agty project folder + the repo-maps standard. | a name → a fully-scaffolded project | maintain-repo set-up-claude-chat-history-graph |
| 🟢 setup-pre-commit | “add pre-commit hooks” · “set up Husky” | Husky pre-commit hooks with lint-staged (Prettier), type-checking, tests. | a JS repo → pre-commit hooks | git-guardrails-claude-code |
| 🟢 git-guardrails-claude-code | “block dangerous git” · “git safety hooks” | Claude Code hooks that block push / reset --hard / clean / branch -D before they execute. | a repo → git-safety hooks | setup-pre-commit |
| 🟢 resolving-merge-conflicts | an in-progress merge/rebase conflict | Walks an in-progress git merge/rebase conflict to resolution. | conflicted tree → resolved merge | — |
Index code + history; look up house facts and model rules.
| Skill | Trigger | What it does | Inputs → Outputs | Related |
|---|---|---|---|---|
| 🔵 graphify | /graphify · any “where is X / how is Y wired” question | Turns any input (code/docs/papers/images) into a persistent knowledge graph with query/path/explain. Query it FIRST when graphify-out/ exists. | a repo/input → graph.json + query answers | plan-explore plan-techspec diagnosing-bugs |
| 🟢 set-up-claude-chat-history-graph | “why did we decide X” · “set up chat-history graph” | Builds + queries a graph of this project's Claude conversations — the strategic record of what was decided. | chat history → strategic-decision graph | new-project finish-session-tts plan-explore |
| 🟢 obsidian-vault | “find/create a note” · Obsidian | Search, create, manage Obsidian notes with wikilinks + index notes. | a query/note → Obsidian vault changes | — |
| 🟢 house-reference | “which env var holds X” · “where does credential Y live” | On-demand reference for Matt's environment — API-keys/services table, file paths, accounts. | a “where/which” question → the house fact | — |
| 🔵 model-help | /model-help · “which model/effort” | Prints the Opus-vs-Sonnet + effort-level decision sheet + session-hygiene rules. | a model question → the decision sheet | — |
| 🔵 chat-context-path | /chat-context-path | Returns the absolute path to the current chat's raw JSONL transcript file on disk. | active session → path to its .jsonl log | — |
From raw fragments to a finished, shippable piece — and out the door.
| Skill | Trigger | What it does | Inputs → Outputs | Related |
|---|---|---|---|---|
| 🟢 writing-fragments | “fragments” · “ideate” · “raw material” | Grilling session that mines you for nuggets (claims, vignettes, sharp sentences) into one raw-material doc. | your head → a fragments doc | writing-shape writing-beats |
| 🟢 writing-shape | “shape these notes into an article” | Conversationally turns a pile of notes/fragments into a publishable article — drafting openings, growing it paragraph by paragraph. | raw material → shaped article | writing-fragments writing-beats edit-article |
| 🟢 writing-beats | “assemble as a narrative” · beat-by-beat | Shapes an article as a journey of beats, choose-your-own-adventure style — you pick the next pivot. | raw material → narrative article | writing-shape writing-fragments |
| 🟣 edit-article | explicit invoke | Edits + improves an article draft — restructure, clarity, tighten prose. | an article draft → a better draft | writing-shape |
| 🔵 tts | /tts <text> · “voiceover” · “read this aloud” | Google Gemini TTS → an MP3 (+ WAV), hosted on Cloudflare, sent to Telegram. | text → MP3 + download link | finish-session-tts cloudflare-staging |
| 🟢 send-to-dani | “send this to Dani” · “küldd el Daninak” | HTML-summarises content, deploys to Cloudflare, emails daniel@clientsflow.hu + Slack. | content → emailed + Slacked HTML link | cloudflare-staging |
The money + outreach pipeline: leads → campaigns → contracts → signatures → invoices.
| Skill | Trigger | What it does | Inputs → Outputs | Related |
|---|---|---|---|---|
| 🟢 gmaps-leads | “scrape Google Maps” · “find local businesses” | Scrapes Google Maps for B2B leads with website enrichment + contact extraction. | a niche+area → enriched lead list | instantly-campaigns |
| 🟢 instantly-campaigns | “create a cold email campaign” · “set up outreach” | Creates Instantly cold-email campaigns with A/B testing. | a lead list + copy → live Instantly campaign | gmaps-leads autoresearch-eval-lab |
| 🟢 autoresearch-eval-lab | “evaluate/rate AI outputs” · “build a rating frontend” | Stands up a human-rating Eval Lab to make an AI sub-skill measurably reliable (frozen set + LLM-judge + keep/discard). | an AI sub-skill → a rating lab + champion | autoresearch-loops |
| 🟢 autoresearch-loops | any AI output iterated with human feedback | Disciplined iteration loop (frozen eval set + explicit metric + experiment log) for refining prompts/templates/copy. | an AI output → a measured iteration loop | autoresearch-eval-lab |
| 🟢 create-contract | “draft a megbízási contract” · engagement contract | Builds a Hungarian engagement contract — a Google Doc review copy + a PandaDoc e-signable, from one master .docx (house formatting). | deal terms → Google Doc + PandaDoc | docuseal-cli szamlazz-invoice |
| 🟢 docuseal-cli | “run docuseal …” in the shell | Manage DocuSeal e-signature workflows from the terminal — templates, send, track, update submitters. | a PDF/DOCX → sent + tracked signing | create-contract docuseal-code |
| 🟢 docuseal-code | “integrate DocuSeal into an app” | DocuSeal dev reference — embed signing forms + builder, full REST API, webhooks, multi-language code examples. | an app → embedded DocuSeal signing | docuseal-cli |
| 🔵 szamlazz-invoice | “állíts ki egy számlát” · “issue an invoice” · “stornózd” | Issues a Hungarian invoice via Számlázz.hu, files the PDF in Drive (link-shared), returns the link. Also STORNO. | vevő + items → számla PDF + Drive link | create-contract |
The skills that build, document, and teach the other skills.
| Skill | Trigger | What it does | Inputs → Outputs | Related |
|---|---|---|---|---|
| 🟢 skill-creator | “create/optimize a skill” · “measure skill performance” | Create / edit / optimize skills + run evals + benchmark + optimize a description for trigger accuracy. | an idea → a registered skill | build-qa-skill writing-great-skills |
| 🟣 writing-great-skills | explicit invoke | Reference for the vocabulary + principles that make a skill predictable. | a skill draft → a sharper skill | skill-creator |
| 🟣 setup-matt-pocock-skills | run once per repo | Configures a repo for the engineering skills — issue tracker, triage labels, domain-doc layout. | a repo → engineering-skills config | triage to-issues review |
| 🟣 teach | explicit invoke | Teaches you a new skill or concept within this workspace. | a topic → a guided lesson | scaffold-exercises |
| 🟢 scaffold-exercises | “scaffold exercises” · “create exercise stubs” | Creates exercise directory structures (sections/problems/solutions/explainers) that pass linting. | a topic → exercise scaffolding | teach |