Skill Claude CodeCodex
How to run the periodic standing-codebase-health audit in the whiteboard repo — fan out auditors across health dimensions, adversarially verify, triage into Tasks/tmp-issues. Use when scheduling or running a periodic review (after a fold, weekly, pre-milestone) for unwired/incomplete features, architecture debt…
Skill Claude CodeCodex
Monitor and triage the POST-PUSH automated-review surface for the whiteboard repo — GitHub Actions CI, CodeRabbit, AccessLint, the WIP app, CodeQL, and Dependabot — into Tasks/tmp-issues. Use after the integrator pushes, when CI/PR checks complete or a review bot comments, to decide what is real and file/fix it.…
Skill Claude CodeCodex
Triage, review, and merge Dependabot dependency-bump PRs + clear Dependabot security alerts for the whiteboard repo. Security-first, semver x ecosystem classification, verify-CI + pnpm audit --prod gate awareness, conflict-cascade-safe batch merging by the single integrator. Adapted for this repo's constraints — NO…
Skill Claude CodeCodex
Make a diagnosis prove it measured what it claims — before reporting a cause, a "not a regression", a mutation check, or a passing verification. Use when debugging a failing test, a CI failure, a flake, or verifying a fix by hand.
Skill Claude CodeCodex
When and how to update the whiteboard project's docs alongside a code change — the trigger→target map, Diátaxis quadrants, doc locations, the honesty rule, and doc-image regeneration. Use when a change has a user-visible effect, alters an API/contract/config, or when triaging a docs gap.
Skill Claude CodeCodex
Build the measuring instrument before changing anything whose effect you cannot see by reading the diff — quality scoreboards (pinned aggregate metrics) and performance benchmarks (pnpm bench, interleaved runs). Use for optimisation, layout/routing quality, heuristic tuning, or any change where "did that help?" is a…
Skill Claude CodeCodex
How to run the whiteboard repo's per-diff change-review gate — args, return contract, and the externalized-criteria mechanism for review.workflow.mjs. Use when running a standalone review gate (pre-merge, pre-un-draft, ad-hoc) or when authoring/tuning the review workflow that dev-loop composes.
Skill Claude CodeCodex
Operates a stack of dependent pull requests in the whiteboard repo with the gh-stack extension — creating a stack across architecture layers, keeping it rebased as main moves, and landing it. Use when a change spans several layers that must merge in order (a shared package under an MCP tool under a web surface), when…
Skill Claude CodeCodex
Drives one whiteboard PR to a green, mergeable, merged state — the order to work a PR event in, which checks actually gate, and the discriminators that separate this PR's failure from the environment's. Use when woken by a PR event (CI failure, review comment, merge conflict), on a PR check-in, or when asked to watch…
Skill Claude CodeCodex
How to pick the nearest test layer in the whiteboard repo (mcp-node / canvas-viewer-node / canvas-viewer-jsdom / canvas-viewer-browser / web-jsdom / web-browser / E2E), when to reach for a property or model-based test instead of an example, and the commands to run each. Use when writing the red test for a change, or…
Skill Claude CodeCodex
Local-private task/ticket management for the whiteboard repo. Issues and notes live in the whiteboard itself (via MCP tools) as OKF Markdown documents. The native Task list tracks in-flight session work. Use when triaging findings, tracking work across dev-loop/review/dogfood-triage/reconcile/plan-initiative, or…
Skill Claude CodeCodex
Produce a before/after figure for a rendering change in this repo — render both versions through the real canvas-render pipeline, compose them side by side, and attach the PNG to the PR. Use whenever a change moves pixels (edge routing, layout, theme, SVG backend) and AGENTS.md's PR Visual Evidence rule applies.
Skill Claude CodeCodex
A skill that standardizes whiteboard MCP and daemon verification. Use it after changing tools, routes, WebSocket behavior, daemon startup, checkpoints, export, or viewport logic when you need to decide what must be restarted and run the right smoke checks. Also use it when the user asks to verify behavior, run smoke…
Skill Claude CodeCodex
Conventions and hard-won gotchas for authoring/maintaining the whiteboard repo's Dynamic Workflows (.claude/workflows/.workflow.mjs) — dev-loop, review, dogfood-triage, reconcile. Use when editing or adding a Workflow script, debugging args not arriving, composing workflows, or running parallel dev-loops.
Skill Claude CodeCodex
How to keep Zod as the single source of truth for any contract that crosses a process boundary in the whiteboard repo (MCP tool input/output, Hono routes, persisted JSON, websocket/daemon payloads). Use when adding or editing an MCP tool, a route response, or a persisted-JSON parser, or when fixing a schema-vs-runtime…
MCP server Claude CodeCodexCursor +2
MCP server + Excalidraw whiteboard UI for AI-assisted diagramming (Claude Code / Codex). Runs locally from the @kamiazya/whiteboard-mcp npm package.
Instructions file CodexOpenCode
Instructions for kamiazya/whiteboard, covering project instructions, test layer selection, property-based / model-based testing (pbt), required workflow and 1. write the red test first.
Instructions file
Instructions for kamiazya/whiteboard, a project described as: An collaborative diagramming MCP tool for AI. Use it to align on specs, architecture, workflows, and visual explanations by drawing together on a shared real-time canvas.
Skill Claude CodeCodex
Audit a whiteboard workspace's documents to find likely-stale or duplicate spatial canvases before creating new ones. Use when a workspace has been in heavy use and you want to check for clutter, or you want a quick sense of what a document actually contains before opening it.
Skill Claude CodeCodex
Evolve a whiteboard diagram collaboratively through structured context gathering, iterative refinement, and fresh-viewer testing. Use when a diagram feels misunderstood, under-explained, or needs tightening together with the user.
Skill Claude CodeCodex
Draw diagrams with your AI agent on a shared JSON Canvas whiteboard. Use it when screen layout, structure, flow, or comparison still feels too ambiguous in text alone. Covers node/edge editing and SVG rendering only — no icon libraries, no other export formats.
Agent Claude Code
name: code-analyzer description: Use this agent when you need to analyze code changes for potential bugs, trace logic flow across multiple files, or investigate suspicious behavior in the codebase. This agent specializes in deep-dive analysis while maintaining a concise summary format to preserve context. Perfect for…
Agent Claude Code
name: file-analyzer description: Use this agent when you need to analyze and summarize file contents, particularly log files or other verbose outputs, to extract key information and reduce context usage for the parent agent. This agent specializes in reading specified files, identifying important patterns, errors, or…
Agent Claude Code
Executes parallel work streams in a git worktree. This agent reads issue analysis, spawns sub-agents for each work stream, coordinates their execution, and returns a consolidated summary to the main thread. Perfect for parallel execution where multiple agents need to work on different parts of the same issue…