Software Development Skills

AGENTS.md maintenance rules for deciding when durable repo guidance should be updated after implementation.

Expert guidance for building AI agents with ToolLoopAgent (AI SDK v6+).

Expert guidance for AI SDK Core: text generation, structured data, tool calling (tool/dynamicTool), MCP integration (createMCPClient, ExperimentalStdioMCPTransport),…

Expert guidance for building chat UIs with AI SDK React hooks. Use when: (1) Building chatbots with useChat hook, (2) Implementing tool UIs with server/client execution, (3) Handling message persistence and stream resumption, (4) Creating generative UI with React components, (5) Integrating with Next.js/Node/Fastify/Nest backends, (6) Using useObject for streaming structured data. Triggers: useChat, useObject, useCompletion, chat UI, chatbot, generative UI, message persistence, @ai-sdk/react, UIMessage, tool parts, streaming UI, toUIMessageStreamResponse.

Architect AWS infrastructure and application integrations with production-grade service selection, boundaries, rollout thinking, and verification. Use when the user asks for AWS architecture, infra design, service selection, or AWS-backed implementation planning. Do not use for repo-specific Nova guidance unless the active repo adapter calls into it.

Batch-grill a plan or design round by round: whole settled frontier per round, recommended answers + close-call scores, ADRs + a glossary via domain-modeling, /pre-mortem handoff on irreversible calls.

Set up browser automation and UI QA for a new repository using playwright-interactive as the primary interactive tool and agent-browser as the secondary CLI smoke-check tool. Use when a user wants to bootstrap browser testing, screenshots, auth-state reuse, or local UI debugging conventions in a fresh repo.

Shared Bun audit/remediation router. Use when auditing a repo for Bun-first correctness, explaining Bun audit findings, listing Bun audit rules, planning safe Bun fixes, applying low-risk remediations, or validating Bun-related changes. This skill delegates to the shared engine in bun-dev/scripts.

Definitive, rule-first Bun development/build/runtime guidance + automation. Use when adopting Bun, migrating a repo from Node.js, auditing/fixing Bun package management (bun.lockb, bun install), optimizing scripts/monorepos (bun run --parallel/--sequential, --workspaces/--filter), configuring Bun + TypeScript, using bun test/build, tuning performance, or deploying Bun workloads/Vercel Functions with the Bun runtime (bunVersion, limitations, Next.js ISR scripts).

Compress natural language memory files (CLAUDE.md, todos, preferences) into caveman format to save input tokens.

Author, validate, and install custom Claude Code subagents as .claude/agents Markdown files. Use when the user asks to create a subagent, custom agent, or .claude/agents role; define a reusable reviewer, explorer, planner, or worker persona; set an agent's model or tool allowlist; or turn a recurring workflow into a delegatable role. Produces Markdown agent files with YAML frontmatter (name, description, model, tools) plus a focused system-prompt body, scoped to the smallest tool surface that does the job (read-only by default). Covers naming, description wording for autonomous delegation, model policy (Opus for reasoning/review, inherit for scans), tool scoping, project vs user scope, and a smoke test. Pairs with fan-out, which delegates to these roles. Not for one-off ad-hoc delegation (use fan-out) or authoring Agent Skills (SKILL.md files).

Architect-level guidance, workflows, and scripts for building agentic coding systems with OpenAI Codex. Use for Codex SDK (@openai/codex-sdk) threads + streaming JSONL events; Codex CLI automation (codex exec, output schema, JSONL, resume); MCP server usage (codex mcp-server) and dynamic tool integration; multi-agent orchestration with OpenAI Agents SDK (handoffs, gating, tracing); durable state, caching, and memory using SQLite; safe-by-default sandbox/approval patterns and execpolicy rules.

Use built-in Codex coordination, retrieval, and utility tools effectively during live sessions. Trigger when work benefits from maintaining a step plan, asking the user 1-3 short multiple-choice questions before risky edits, inspecting a local image from disk, searching or browsing the web for current or source-backed information, discovering additional MCP tools or plugins, or running independent developer-tool calls in parallel.

Deep library, dependency, framework, and API documentation research using Context7 MCP tools only. Use when you need up-to-date docs, version-specific API reference, migration guidance, or to verify code against official documentation. Triggers on requests like “look up docs”, “API reference”, “how do I use X”, “latest/current”, “deprecated”, “breaking changes”, “best practices”, or whenever correctness depends on primary documentation.

Audit Convex—schema, security, runtime edges, migrations, function risk. Triggers—review, contract, remediate. Not greenfield spec (convex-feature-spec).

Researches a Convex component against Signr's live backend graph, asks scored one-at-a-time design questions, and produces a reusable adoption or rejection package with PLAN.md and CODEXFULLPROMPT.md.

Generate a complete Convex-first feature specification for a repo, including data model, API surface, rollout, and verification. Use when the user wants a new feature plan or architecture spec centered on Convex contracts. Do not use for audits of the existing implementation; use convex-audit for that.

Audit Dash apps for callback hazards, state flow risks, layout and accessibility issues, and Dash-specific UX regressions. Use when the user asks for a Dash callback review, Dash UI audit, Dash web-interface audit, or a read-first remediation plan for a Dash app. Do not use for generic React or Next.js UI review.

Performs comprehensive, multi-layered research on any topic with structured analysis and synthesis of information from multiple sources.

Audit a repository, route, screen, component, or 3D scene for motion quality — design-token consistency, R3F/three.js polish, Reanimated native motion, interaction physicality, frame-rate/draw-call risk, reduced-motion coverage, accessibility, and missing hallmark-motion opportunities. Use when asked to audit, review, critique, find gaps, check performance, verify reduced motion, or assess whether UI animation feels premium. Returns a prioritized punch list. For building/implementing motion route to design-motion-system, expo-motion, web-three-r3f, r3f-scene-polish, or gsap.

Cross-stack motion & design-system director. Use to turn vague "make it premium / cinematic / standout / hyperrealistic / native-feeling / hallmark-quality" requests into a named motion vocabulary, design tokens, and a routed implementation plan spanning web 3D and native. Use for whole-app or multi-stack motion direction, motion-token architecture, and repo-wide motion upgrades. Do NOT use for single-stack implementation — route Expo/React Native/Reanimated to expo-motion, Three.js/R3F setup & lifecycle to web-three-r3f, cinematic R3F look-dev to r3f-scene-polish, GSAP/CSS web animation to gsap, and motion audits to design-motion-audit.

Dash Mantine Components (DMC) best practices and performance optimization guidelines. Use this skill when writing, reviewing, or refactoring Dash apps with DMC to ensure optimal patterns. Triggers on: DMC components, Mantine styling, Dash callbacks, dcc.Store, theming, or performance optimization in Plotly Dash applications.

Expert guidance for building Dash applications with Dash Mantine Components (DMC) v2.4.0. Use when creating dashboards, forms, data visualization apps with DMC. Covers: MantineProvider theming, style props (m, p, c, bg, w, h), Styles API, callbacks (basic, pattern-matching ALL/MATCH/ALLSMALLER, clientside, background), multi-page apps with Dash Pages, charts (LineChart, BarChart, DonutChart), date pickers, modals, and all 90+ components. Triggers on: dash-mantine-components, DMC, MantineProvider, dmc.Button, dmc.Select, dmc.Modal, dmc.BarChart, Mantine theme, Dash UI components, Dash callbacks, multi-page Dash app, pattern-matching callbacks, clientside callbacks, AppShell.

SOTA Docker/Compose architecture, implementation, refactor, and security hardening. Use when working on containerization tasks such as creating or rewriting Dockerfiles, docker-compose files, buildx/bake configs, .dockerignore, and CI pipelines for build/test/scan/publish; auditing existing container setups for security, correctness, size/perf, and best practices (least privilege, non-root, minimal images, pinned base images, BuildKit secrets, healthchecks); debugging Docker build/run issues; or designing dev vs prod compose workflows across services (DB/cache/queues) with correct networking, volumes, secrets, and resource limits.