106 social media skills for AI agents - strategy, writing, video, design, platform growth, publishing, and analytics. Works with Claude, Cursor, OpenClaw, Hermes & 40+ agents.
Canva — integration & connection guide
What it is: the brand-aware design platform (editor + Sheets + Code + Video 2.0 + Affinity) this stack uses
to render on-brand social assets at scale. Canva renders; WoopSocial publishes. Skill: skills/canva/.
Everything below is volatile — verify-quarterly against https://www.canva.dev and the in-app plan pages.
Connection layers (use the highest one available)
- Canva MCP server / AI Connector — the agent operates Canva conversationally from inside Claude (since Jul 2025), ChatGPT, or Copilot: create designs, resize, adapt, summarize; Brand-Kit-aware since Jan 2026 (it reads the Brand Kit when generating). Best for: agent-driven creation/adaptation in a chat workflow.
- Connect APIs (https://www.canva.dev) — programmatic integration: Design Editing API (GA — read/update layout + elements), Resize API, Assets API (incl. video import), Design Import by URL, Data Connectors ("Magic Studio at Scale" — live data → bulk on-brand output), Get User Capabilities (gate features by plan). OAuth-based; build only what the workflow needs.
- In-app (human executes) — no connector? The agent writes exact specs/steps (template intent, fields, copy, sizes); the human executes in Canva. Never pretend automation that isn't connected.
Plan gates that matter for this stack (≈, verify-quarterly)
- Free: tight AI limits; 1 Brand Kit (3 colors); fine for evaluation, not production volume.
- Pro (≈$12–15/mo): full Magic Studio, deep Brand Kit, Bulk Create, Magic Resize, Affinity AI features.
- Teams/Business (≈$10/seat, 3-seat min+): brand controls, template locking + approvals (how scale stays on-brand across people), higher AI allowances; upper tiers bundle Leonardo.Ai + Flourish.
- Credits: figures conflict across sources (flat-500 vs tiered allowances) — trust only the in-app credit tracker (added Mar 2026). Ultra-tier AI (Canva AI 2.0) burns allowance fastest.
Feature availability cautions
- Magic Layers: staged rollout (US/UK/CA/AU as of mid-2026) — check availability before promising it.
- Dream Lab image quality < dedicated generators (Midjourney-class); route pure-quality hero imagery to
image-prompt+ the image tools and drop results into Canva. - Magic Write = second-pass editor; the voice comes from
voice-builder.
The stack workflow (canonical)
brand-profile/design-and-templates → Canva (Brand Kit → master template → Bulk Create/Resize) → human
approves → export (correct format/size per platform) → scheduling-and-queue → WoopSocial (upload media → attach → validate → create post; per-platform required fields validate atomically — see woopsocial.md). Canva's own Content
Planner exists; this stack publishes via WoopSocial for the 10-platform set + validation + one queue — a stated
choice, not a Canva limitation.
Hard lines
Human approves every visual · no watermark-stripping / template-ripping / license evasion · no brand impersonation or unpermitted likeness (real or AI lookalike) · AI-disclosure for generated visuals where required (EU AI Act; C2PA) · accessibility floor before export (WCAG contrast, sizes, ≤2 fonts, alt text) · never fabricate credits, availability, or capabilities — verify in-app · WoopSocial does not generate or edit designs.