Claude Skills by byjunyoung

Generates and re-syncs flow arrows and their labels on a Figma design page. Reproduces in code what Autoflow-style plugins do — edge-midpoint anchors, right-angle elbows, branch trunks, element anchors, line styles that carry meaning. Flow information is stored in the node name (`source --> target`), so the whole page re-syncs with one command after frames move. Triggers - "/fig:arrows", "draw the flow arrows", "connect these screens", "re-sync the arrows", "화살표 그려줘", "흐름 화살표 추가", "화살표 sync",...

Applies a Figma design to real code in a frontend repo — porting changed parts into existing code with minimal edits, or implementing screens and state variants that exist only in the design. Reads the repo's own naming, tokens, and state-management conventions first so it never overwrites them, checks both by numbers and by screenshot comparison, then opens a branch and a PR. Triggers - "/fig:code", "apply this design to the code", "implement this screen in the frontend", "시안 코드에 반영해줘", "피그마...

Measures a team slide template into local deck assets — template-spec.md, template.js, and reference images. Canvas, type scale, colors, and the archetype catalog are read from actual nodes; anything the sample is too thin to settle is left empty rather than guessed. Run once before using /fig:deck in a new environment, or again after the template is revised. Triggers - "/fig:deck-setup", "extract the slide template", "set up the deck assets", "덱 템플릿 뽑아줘", "발표 템플릿 추출", "템플릿 다시 추출".

Turns a document, a dataset, or a spoken brief into a Figma Slides deck. Reads the source, agrees an outline before writing anything, then builds slides from the team template's archetypes and audits overlap and wording before handing back. Research reports, plans, retrospectives, proposals — it does not care which. Template values come from the local assets /fig:deck-setup produced. Triggers - "/fig:deck", "make this into a deck", "turn this doc into slides", "슬라이드로 만들어줘", "발표 자료 만들어줘", "덱 만...

Compares AS-IS and TO-BE designs, marks the changed elements with native Figma Dev Mode annotations, and writes up the linked task doc with a Figma link, an AS-IS/TO-BE table, and a scope callout. The tracker comes from the task_tracker section of figma-conventions.yaml and can be notion, github, or none; with none it stops at the Figma annotations and emits the table as markdown. AS-IS may sit in another section of the same page or on a different page entirely, so the source is settled first...

Hands finished sections to engineering. Runs /fig:lint as the gate, lets the person pick which passing sections go, hands over the section links, and writes one line into the task doc where a tracker is configured. Marking them Ready for dev in Figma is off by default — use_figma cannot write a dev status. Nothing else is touched. Triggers - "/fig:handoff", "hand this off", "mark these ready for dev", "send this to engineering", "개발 넘겨줘", "핸드오프 해줘", "Ready for dev 표시해줘", "이 섹션 개발팀에 넘겨".

Audits a Figma design page read-only — frame membership and bounds, section overlap, naming, arrow geometry and entry direction and pass-through, coverage orphans, [state] dashed links, variants stacked on one another inside a component set, and component default residue carried over by duplication. Reports violations and writes nothing. Rules come from figma-conventions.yaml and the audit code lives in one place under ${CLAUDE_PLUGIN_ROOT}/_common/scripts. It is the single gate that absorbs ...

Prepares a Figma design page before the real work — normalizing frame names to `[screen]-[state]`, creating and placing feature-level sections, and reading screen content to find missing cases (state variants, interaction results) and stub them as placeholder frames. Includes the absorb workflow that works around reparenting limits with local font files. It handles renumbering and handoff URLs; it does not verify or report violations — both status checks and completion checks belong to /fig:l...

Rebuilds a Figma design as code so that input, validation, state, and branching actually work in a single HTML file. Not a click-through screenshot demo — something you press to check the UX and get a feel for the implementation spec. The pipeline runs as settle scope by interview, read Figma faithfully (structure, exact labels, visual tokens), build as vanilla single-file HTML, then verify real behavior in a browser. Triggers - "/fig:proto", "make a working prototype", "make this design actu...

Audits a feature that has shipped to a dev or staging server against the plan of record — spec entries and the Figma design. One mode stops once the checklist is drawn; the other clicks through the browser and files a defect report. Defects are written as reproduction steps, expected, actual, and evidence (request results, screens) so they can be handed to engineering as they are. Triggers - "/fig:qa", "QA this", "check what is on the dev server", "file the defects", "QA 해줘", "개발서버 확인해줘", "결함...

Takes a Figma file URL, lists its pages, and collects every top-level frame on the page you pick. Metadata and screenshots are read in parallel and laid out as a markdown tree. Gives you the whole structure from a URL alone, without selecting frames one by one in the desktop app. Triggers - "/fig:read", "read this Figma file", "list every frame", "피그마 파일 읽어줘", "프레임 전부 뽑아줘", "전체 구조 보여줘".

Reads an unfamiliar Figma file and works backwards to its conventions, drafting a figma-conventions.yaml. Page roles, naming patterns, state suffixes, spacing tokens, section style, and arrow style are observed from actual nodes and taken as the dominant value; anything with too few samples or a split vote is left null rather than guessed, then filled in by interview. It finishes by running /fig:lint and judging the draft by its false-positive rate. Run this first when opening these skills on...

Audits whether changes finished on working and update pages actually made it into the canonical page, applies whatever did not, then archives the working copies. This is the skill that executes the apply-after-release step — audit with zero writes, then apply, then archive, each behind its own preview and go gate. Frame names are identical on both sides so comparing names settles nothing; the call is made on three signals together — text diff, frame height, and component master. Triggers - "/...

Audits whether the colors on a Figma frame or page are bound to design system tokens, and maps and binds hardcoded colors to the right token. Derives token-to-hex from bindings already in the file so every mapping has evidence, grades the results, and auto-proposes only the safe ones. Lint mode (check only) and bind mode are separate. Which library and which token groups come from the design_system section of figma-conventions.yaml; on auto it detects the library the file is connected to. Tri...

Turns a period of daily logs into accomplishment statements you can reuse in a self-review, a promotion case, or a CV. Stitches entries by tracker url so one piece of work spanning weeks becomes one item, then asks only what the files cannot know — your role, what the result can be measured by, what you learned — and writes the answers into an accumulating highlights document. It never rates your work for you and never supplies a number you did not give. Triggers - "/pm:log-review", "turn las...

Writes one day's work log as a file — what moved in the tracker, what was on the calendar, what was decided, what other people said about your work, and links to the evidence. Built to run unattended on a schedule, so it never asks and never invents; a source it cannot reach is reported as missing rather than filled in. It also looks back a few days and fills the gaps an earlier run left. Turning these files into accomplishment statements is /pm:log-review. Triggers - explicit invocation only...

Writes a product requirements document against a format, or fills out an existing one. Gathers and analyzes code repos, docs, designs, and chat history first, then proposes recommended answers so the interview can be settled with a single "go". Before writing anything it verifies read-only for vague wording, empty definitions, and engineering terms that do not belong in a product doc, then reads the entries against each other for roles nobody defined, entries that contradict, cases nobody wro...

Reads the schemas of your document tool and your issue tracker and drafts pm-conventions.yaml from them. Property names, select options, labels, board field ids and user mappings are read from the live schema rather than guessed, and the ids nobody can find by hand — board node id, custom field ids, single-select option ids, issue-type ids — are queried for you. Anything the schema does not settle is left null and asked about rather than filled in. Run this first when opening these skills on ...

Turns a request source — a chat thread, a page, a spoken agreement — into a task record's context table. Statements from the source are sorted into named rows, and what is fact, what is your inference, and what nobody has decided yet are labelled apart rather than blended. Creates a new record where none exists, or fills only the empty cells of one that does. Filing it in the engineering tracker belongs to /pm:task-publish. Triggers - "/pm:task-draft", "draft a task from this thread", "fill i...

Files one task record as a ticket in the engineering tracker, or updates the ticket already there. The ticket carries a summary and links rather than a copy of the body, and it carries the two sections that decide when the work is finished — the done conditions and the QA checklist — drafted from the spec and the design rather than left as a placeholder. Missing fields are filled by one grouped interview, the parent is resolved before writing, and the link back into the record is what matchin...

Reconciles a planning-side task list against the engineering tracker it mirrors. Reads both, diagnoses the ways they have drifted apart — unfiled, duplicated, wrong parent, broken link, resurrected, field mismatch — and then proposes, waits for approval, and only then writes. Nothing is written automatically. Where the read cannot be exhaustive it says so in the coverage line rather than implying it saw everything. Filing or updating one task belongs to /pm:task-publish. Triggers - "/pm:task-...