AI Agent Workflow Collection

Design stable, hard-to-misuse interfaces — REST/GraphQL endpoints, module boundaries, type contracts, component props, anything where one piece of code talks to another. Use at design time, before implementing the surface. Triggers: 'design this API', 'API contract', 'endpoint design', 'module boundary', 'interface design', 'should this be one endpoint or two', 'how should this API look'.

Inspect and debug Airflow on Astronomer (Astro) deployments — fetch DAG runs, task instance logs, container logs, env vars, and deployment state without installing an MCP plugin. Use when the user mentions Astro/Astronomer, asks about DAG runs or task logs on staging/prod, says 'check the deployment', references astro deployment, make airflow, an Astro deployment ID, or a .astronomer.run URL. Pairs the official astro CLI for platform ops with direct Airflow REST API calls for DAG-level data.

Manage named persistent browser profiles (cookies, localStorage, IndexedDB, cache) that both you and Claude can use against the SAME Chrome window without collisions. Each profile lives in its own --user-data-dir and exposes a deterministic --remote-debugging-port. Use when the user says "open profile X", "attach to profile X", "log me into staging once and reuse it", or asks to test frontend flows with persistent auth across runs. Pairs with the vd:browser skill (browse env local --auto-connect) for the CDP attach step.

Migrate from claudekit (ck) to the vd-cli control plane — install clean-room hooks, convert .ck.json to .vd.json, audit CK→VD env consumers, and move a repo's plans/ artifacts into feature-first .workbench folders under /.workbench/features/ /{plans,reports,journals,visuals,state}. Use when the user says 'cktovd', 'migrate to vd', 'migrate plans to .workbench', 'enable the .workbench umbrella', or 'switch this repo off claudekit'.

Build and maintain production TypeScript CLIs. This skill should be used whenever work touches commands, packaging, CI/CD, npm publishing, docs, providers, or agent-friendly output.

Build evidence-backed exploratory analysis and GitHub-renderable notebook reports in /Users/vanducng/git/work/cnb/cnb-ds-eda. Use when the user asks for CNB EDA notebooks, Snowflake-backed analysis, Retell or Transfer AI investigations, data-contract-derived report logic, notebook refreshes, or report artifacts under notebooks/.

Create and operate CloudNativePG (CNPG) Postgres databases on Kubernetes the GitOps/Flux way — on managed cloud (GKE + GCS via Workload Identity) OR self-hosted (K3s/bare-metal + any S3-compatible store via a credentials secret). Covers Cluster + ScheduledBackup manifests, barman WAL archiving, pgvector, PITR, prod→dev clones, and the NetworkPolicies a default-deny cluster needs. Use when provisioning a new app database, cloning prod into dev, enabling pgvector, wiring backups/PITR, writing CNPG NetworkPolicies, or debugging the silent "WAL archiving failed → PVC fills → Postgres CrashLoop → app can't read data" chain on CloudNativePG.

Run deterministic multi-step agent workflows in Codex — fan-out, parallel, schema-validated structured output — via the codex-workflow MCP orchestrator (runworkflow), with native in-chat fallbacks. Use when you want Claude-Code-Workflow-style orchestration in Codex: 'run a workflow', 'fan out subagents', 'review each file in parallel', 'orchestrate these steps'. Triggers: 'codex-workflow', 'runworkflow', 'orchestrate in codex'.

Audit and reclaim disk space on macOS. Discovers cache hogs, dev caches, dead app data, old Downloads, dup installers, and obsolete container/VM images. Use when user says 'clean disk', 'free up space', 'computer-clean', 'cleanup mac', 'disk full', or shows ≥85% disk usage.

Design and review production database schemas, data models, ERDs, migration plans, and storage patterns. Use for OLTP schema design, OLAP/star-schema modeling, fact/dimension tables, indexes, constraints, partitioning, multi-tenant data models, CSV/JSON-to-table design, schema review, migration risk review, and database design documents across PostgreSQL, MySQL, SQLite/D1, BigQuery, and MongoDB. Does not execute DDL by default.

Scaffold and build a full-stack web app: FastAPI backend (Python, uv, SQLModel, Postgres, Alembic, JWT + Google OAuth, boto3/S3) + React frontend (Vite, TypeScript, shadcn/ui + Tailwind, TanStack Router/Query/Table, Zod, Axios), wired with Docker Compose. Use this skill whenever the user wants to spin up, bootstrap, create, or design a new full-stack webapp; an API-first backend + SPA frontend; an admin/portal/dashboard app; file upload + S3; RBAC / role-based auth with seeded test users; local docker dev; or asks for a 'FastAPI + React' / 'Python + React' project. Runs mockup-first: marketing-design (brand/logo raster) + opendesign (HTML page mockups) before code, then ports the design to Tailwind/shadcn. Covers project structure, local setup, auth/RBAC, S3 uploads, and the gotchas that break these stacks.

Local HTTP server that renders markdown, code, text, JSON, CSV/TSV/XLSX tables, HTML, PDF, images, video, and audio in the browser. Persistent left tree sidebar with vim keybindings (j/k/h/l/gg/G/Enter/o), filter (/), theme toggle (T), and folder open/collapse persistence per tab. Markdown dispatches to a novel-theme reader (Mermaid, plan nav, ToC scroll-spy); code/text dispatches to a highlight.js view with line numbers + copy button; CSV/TSV/XLSX dispatches to a tabular view; HTML renders in a sandboxed iframe; PDF dispatches to a custom minimal viewer built on the pdfjs-dist PDFViewer component (toolbar: prev/next, page input, zoom, search, download); images/video/audio get a gallery + single-view with Range-aware streaming. One server, one port, one CLI.

My curated Go stack reference — opinionated guide to Sam Berthe's libraries (lo Lodash helpers, oops structured errors, do v2 DI, mo monads, slog logging ecosystem, hot in-memory cache pre-1.0, ro reactive streams pre-1.0). Use when writing or reviewing Go code that imports github.com/samber/, when deciding between these libraries and the stdlib, when adopting or upgrading any of them, or when scaffolding a new Go service and considering this ecosystem.

Set up an isolated git worktree for a Laravel project served by Laravel Herd (Pro-aware). Composes vd:worktree for the worktree/.env/port mechanics, then adds the Herd layer: link + scheme-correct site (HTTPS via herd secure when the source is secured), .env rewrite (APPURL, SESSIONDOMAIN, Sanctum — only when present), opt-in per-worktree database isolation, Vite TLS/CORS, then hands finishing to vd:ship / vd:git. Triggers: 'herd-worktree', 'herd worktree', 'laravel herd worktree', 'isolate a Laravel feature branch', 'work on this Laravel branch with Herd'.

Review code/diffs/PRs with the owned miucr CLI (miu-cr, a pure-Go AI code reviewer). Use when asked to review staged changes, a commit, a ref range, or a GitHub PR; to run/parse a gated review; compare reviewer quality with eval; to drive reviews over MCP; or to run the serve webhook/poll daemon or GitHub Action. Output is the stable miucr.cli/v1 JSON envelope; parse it, don't grep prose.

Query, inspect, and manage saved database connections through the Go miudb CLI. Use when the user asks to run SQL, list schemas, add native connections, smoke-test connections, inspect tunnel-backed databases, or produce agent-readable JSON from SQLite, Postgres, MySQL, Snowflake, or BigQuery.

Generate polished single-file HTML design artifacts using the nexu-io/open-design catalog of production skills and brand-grade design systems. Use when the user asks to design, mock up, prototype, draft, build, or render a visual web artifact, chooses a brand/style direction, requests a deck/PPT, hero, section, component, marketing page, dashboard, email, doc page, or GitHub README/repo banner. Auto-opens the result for review. Produces editable HTML/CSS; use vd:marketing-design for AI-generated raster brand imagery.

Autonomous metric-optimization loop — run N bounded iterations against a mechanical metric, commit each attempt, auto keep/discard on the number, revert regressions. Use to improve a measurable metric: test coverage, bundle size, lint/type-error count, p95 latency, LOC. Triggers: 'drive coverage up', 'reduce bundle size', 'get lint errors to zero', 'optimize over iterations', 'keep/discard loop'.

Migrate Python projects to idiomatic Go end-to-end. Branches into 6 project-type playbooks (CLI, TUI, HTTP backend, data pipeline, async worker, library) with the right stack defaults (Gin, pgx, sqlc, slog, etc.) and pinned library versions. Default strategy: LLM module-by-module rewrite with golden-file parity tests; --strangler for live-traffic gradual cutover; --spec-first for OpenAPI/proto-driven regeneration. Use when porting a Python codebase to Go, when scaffolding a Go rewrite of a Python service, or when generating CLAUDE.md/MIGRATION.md for an AI-driven migration.

Mine your Claude Code sessions and code-review history for the corrections you keep repeating, cluster them, adversarially verify each candidate, and distill the survivors into CLAUDE.md rules — or check that existing rules still hold. Uses a dynamic Workflow harness with one verifier per rule to avoid false positives. Use when the user says 'mine my sessions for rules', 'what corrections do I keep making', 'turn my repeated feedback into CLAUDE.md rules', 'audit my CLAUDE.md rules', or 'why does Claude keep missing X'.

Create polished self-contained HTML showcase pages and section screenshots for demos, launch posts, social media, article illustrations, and visual presentations. Use when the user wants to turn a project, feature, case study, markdown brief, or prompt into a bilingual scrollable showcase with reusable promo images.

Operate the official Superwhisper CLI and MCP to search or read dictation history, build standups and commitment reviews, compare raw and processed transcripts, inspect usage, and manage vocabulary or snippets with approval. Use when the user asks what they said about a topic, wants recent voice notes summarized, needs Superwhisper transcription diagnosis, or asks to inspect, export, or maintain Superwhisper data.

Draw reliable, well-aligned text/ASCII architecture diagrams with nested boxes, arrows, and connectors. Use when user asks to draw/create text diagrams, ASCII art diagrams, box diagrams, or architecture diagrams in plain text. For rendered image artifacts (PNG/SVG), use vd:diagram instead.

Operate the user's tldraw offline canvas app, including open .tldraw or .tldr files. Use whenever a task involves inspecting, editing, arranging, connecting, linting, or scripting a tldraw Desktop canvas.