figma-maxxing

Single owner of the connection between Figma Desktop and the figma-console MCP Desktop Bridge plugin, alias-agnostic, across every project. Use on any intent to open, activate, restart or check Figma or the bridge, and before any figma-console tool call. Trigger phrases: "open the Figma file", "is the bridge connected?", "restart the bridge", "abre o figma", "liga a bridge". Probes state, takes the minimum action, verifies with figma_get_status and escalates at most 4 times. Always Figma Desktop, never the browser. The automation is macOS only.

Read-only knowledge base of Figma canon for AI agents: a routing table of on-demand references (Plugin API core, data and anomalies, write atomicity, auto layout, inspection, handoff format, naming, state coverage, quality rubric, AI slop signatures, security, rate limits) plus hard rules learned in production (the canvas mutates under you, null-guard before operating, hex versus token). Use on any mention of Figma, a node id, a design system, a token, a component or auto layout, and before any figma_execute call. It informs writes and never performs one.

Draws a click-flow overlay on a section of laid-out Figma screens for static handoff: a dot on each clickable element, an elbow line with rounded corners and an open arrow at the destination screen. figma.createConnector is not a function in Figma Design files, so each connector is emulated with one SVG import (hardened recipe in references/draw-overlay.js). Use when the user asks for a "click flow", "flow arrows", "wire screens", "turn this into a flow", or in Portuguese "vira em fluxo", "setas do fluxo".

Comment-to-fix pipeline for Figma. Pulls the WHOLE set of open comments (REST API or figma_get_comments with FIGMA_ACCESS_TOKEN, because the Plugin API cannot read comments), clusters them by frame, fixes each one on the screen where its pin sits, and builds an evidence package. Nothing can resolve a comment through the API, so the human resolves the threads. Use when the user pastes a stakeholder comment, even a single one, or says "fix what they commented", "address the review comments", "apply the feedback left in the file", "corrige o que comentaram".

Pre-ship gate for a Figma handoff. Runs 17 checks before a section is declared ready for developers: action completeness (every drawn action has a designed outcome, and inverse pairs such as add and remove both exist), annotation layout and contrast, dense spec cards with no AI stuffing, a hand-crafted surface with no trace of the process, and proof at the right scale. Run figma-slop-check first. Use when the user says "ready for handoff", "dev-tag", "release section", "spec card", "run the handoff gate", or in Portuguese "pronto pra handoff".

First move on a Figma file or URL that is new to the session. Builds the map of the file (pages, canonical screens, design system source, WIP zones) through the figma-console Desktop Bridge, the primary read path because the official Figma MCP server has a plan-based call cap, and saves it to docs/figma-map.md (or the agent's memory) so later sessions do not walk the file again. Read-only on the canvas. Use when a figma.com URL is pasted with no task attached, or on "what is in this Figma file", "orient on this file", "map this file", "o que tem nesse figma".

Mandatory read-only gate that runs before any write to a Figma file, through figma_execute (figma-console MCP) or use_figma (official Figma MCP). Runs 8 checks in order (live connection, confirmed target node-id, metadata, variables, design system and icon search, canon references, state coverage, file lock) and either approves the write or returns a fix list. It never mutates the canvas. Also audits cross-screen flow: orphan screens, CTAs without a destination, dead ends. Use when the user asks to create, edit, build or instantiate anything in Figma, or asks "audit the flow", "audita o fluxo", "tem tela órfã?", "esse botão leva pra onde?".

Quality gate that runs after a write to a Figma file and before figma-handoff-gate, before any frame is called done. Two lenses. Slop: does it look machine made (overlap, AI-stuffed cards, status emoji creep, hype copy, names and canvas notes that read as generated). Rigor: is it precise (values bound to variables and styles, sizes on the project's scales, components intact, auto layout sized right, one wording per concept, screenshot proof). It reads and reports: PASS, or a numbered punch list by node id. Use when the user asks "is this ready?", "final review", "audit this screen", "ship it", "tá pronto?", "pode fechar", "audita".