modal-sandboxes

Complete Tailwind CSS v4 expertise with 2025 best practices. PROACTIVELY activate for: (1) any Tailwind task, (2) v4 CSS-first config (@theme, @utility, @custom-variant), (3) responsive design and dark mode, (4) performance and JIT, (5) plugins (typography, forms), (6) framework integration (React/Vue/Next/Vite), (7) v3-to-v4 migration, (8) custom utility creation, (9) debugging, (10) production build optimization. Provides: v4.1 features, PostCSS and Vite setup, dark-mode strategies, animation utilities, VS Code/Prettier config.

12 skills · 4 commands · 1 agent

Comprehensive Python 3.13+ expertise covering modern Python, async, types, performance, and cloud deployment. PROACTIVELY activate for: (1) any Python task, (2) 3.13+ features (free-threading, JIT), (3) asyncio and concurrency, (4) type hints with mypy/pyright, (5) package management (uv/pip/poetry), (6) project layout, (7) FastAPI, (8) Pydantic, (9) pytest, (10) Cloudflare Workers/Containers Python, (11) GitHub Actions workflows, (12) profiling, (13) OWASP security, (14) Ruff linting. Provides: production patterns, memory profiling, dataclass vs Pydantic guidance.

13 skills · 2 commands · 1 agent

Expert bash/shell scripting across all platforms with Bash 5.3 and 2025 security-first practices. PROACTIVELY activate for: (1) any bash/shell task, (2) Bash 5.3 features (in-shell substitution, BASH_TRAPSIG), (3) security-first patterns (input validation, HISTFILE), (4) ShellCheck v0.11 validation, (5) CI/CD and container-aware scripting, (6) debugging and profiling. Provides: Bash 5.3 reference, ShellCheck rules, Google Shell Style, POSIX.1-2024 compliance, cross-platform patterns, and cloud/CI helpers.

10 skills · 4 commands · 1 agent

Documentation diagnostic, Markdown style, and ADR expert. Diagnostic-first: refuses to write ADRs for non-decisions and routes the rest to the right home (RFC / Diátaxis / runbook / code comment). Co-drafts ADRs with zero-hallucination discovery, scripted push-back, line-by-line audit of legacy ADRs, and canonical-C4 LikeC4 diagrams. Folder-level audits classify every doc as KEEP / MERGE / REWRITE / DELETE / MOVE / BACKFILL-ADR — the last surfaces past architectural decisions that were never recorded (shipped commits, migrations, retired vendors), retroactively recorded via adr-backfill with a mandatory honesty clause and two-locator evidence. Two-pass Markdown lint over any .md file: canonical syntax (Markdown Guide) plus opinionated style (Google Markdown style guide), per-finding with rule citations.

8 skills · 6 commands · 1 agent

Expert fal.ai generative media platform with 600+ AI models (FLUX.2, GPT-Image 1.5, Veo 3.1, Sora 2, Kling 2.6, Recraft V3), client libraries, and serverless deployment. PROACTIVELY activate for: (1) any fal.ai task, (2) image generation, (3) FLUX Kontext editing, (4) video generation, (5) native audio in video, (6) image-to-video pipelines, (7) real-time WebSocket streaming, (8) queue-based execution, (9) serverless deployment (fal.App), (10) GPU compute (T4-B200), (11) JS/TS and Python integration, (12) storage and volumes, (13) cost optimization. Provides: model catalog, client patterns, webhooks, rate limiting.

10 skills · 4 commands · 1 agent

Core FFmpeg encoding, transcoding, diagnostics, audio, subtitles, analysis, restoration, codec selection, and hardware acceleration. PROACTIVELY activate for FFmpeg command syntax, H.264/H.265/AV1/VVC encoding, NVENC/QSV/VAAPI/Vulkan, loudnorm/audio work, subtitle burn-in/extraction, ffprobe analysis, deinterlacing, denoise, stabilization, and quality troubleshooting.

10 skills · 3 commands · 1 agent

Tenant-isolated AI teams that plan work, coordinate tasks, recall approved context, and return evidence-backed results

6 skills · 6 commands · 4 agents · 1 MCP server

Agent Skills for Together AI platform — inference, training, embeddings, audio, video, images, function calling, and infrastructure. Covers serverless chat completions, image/video generation, fine-tuning, batch inference, evaluations, sandboxes, dedicated endpoints, and GPU clusters.

Context window saver for 10 coding agents. Local and free: sandboxed code execution, FTS5 search, session continuity.

3 skills · 1 MCP server

Opt-in, outside every domain bundle: orchestrate opencode workers from Claude Code to cut token costs, with single workers or parallel scout/coder/tester swarms on free or open-source models and worktree isolation. Requires the third-party opencode CLI and an opencode account; by default worker prompts and code go to opencode's hosted service (point --model at a local model to keep code on your machine), so install it only if you accept that data flow.

1 skill

GPU cloud and compute orchestration including Modal, Lambda Labs, and SkyPilot. Use when deploying training jobs or managing GPU resources.

98 skills

Skills plus explorer and reviewer subagents for building, scaffolding, debugging, understanding, and auditing applications with the Atomic Agents Python framework. Progressive-disclosure references for schemas, agents, tools, context providers, prompts, orchestration, memory, hooks, providers, project structure, and testing — plus a troubleshoot diagnostic, isolated-context subagents for codebase mapping and code review, and a new-app scaffolder.

7 skills · 2 agents