NVIDIA's runtime safety framework for LLM applications. Features jailbreak detection, input/output validation, fact-checking, hallucination detection, PII filtering, toxicity detection. Uses Colang 2.0 DSL for programmable rails. Production-ready, runs on T4 GPU.
llamaguard
llamaguard is an open-source backend skill for Claude Code and compatible agents, published by Orchestra-Research. Its author describes it as: “Meta's 7-8B specialized moderation model for LLM input/output filtering. 6 safety categories - violence/hate, sexual content, weapons, substances, self-harm, criminal planning. 94-95% accuracy. Deploy with vLLM, Huggi…”. The project has 12k stars on GitHub and is available under the MIT license. Add it to your setup with `/plugin marketplace add Orchestra-Research/AI-Research-SKILLs`.
What llamaguard does
LlamaGuard is a 7-8B parameter model specialized for content safety classification.
Installation
Add llamaguard to your agent with:
/plugin marketplace add Orchestra-Research/AI-Research-SKILLsAlways review a skill's source before installing it. This command comes from the skill's public repository; the linked repo is the source of truth for exact setup steps.
What's inside
The SKILL.md for llamaguard is organised into these sections:
- Quick start
- Common workflows
- Workflow 1: Input filtering (prompt moderation)
- Workflow 2: Output filtering (response moderation)
- Workflow 3: vLLM deployment (fast inference)
- Workflow 4: API endpoint (FastAPI)
- Workflow 5: NeMo Guardrails integration
- When to use vs alternatives
- Common issues
- Advanced topics
- Hardware requirements
- Resources
When to use it
Reach for llamaguard when you want backend help from your agent without writing the same instructions every session. Load the skill and the agent picks it up automatically for relevant tasks.
Strengths
- Clear MIT license — safe to read and adapt
- Ships in Orchestra-Research/AI-Research-SKILLs, an established project with 11,807 GitHub stars
- Actively maintained (recent commits)
Topics
Frequently asked questions
- What does llamaguard do?
- Meta's 7-8B specialized moderation model for LLM input/output filtering. 6 safety categories - violence/hate, sexual content, weapons, substances, self-harm, criminal planning. 94-95% accuracy. Deploy with vLLM, HuggingFace, Sagemaker. Integrates with NeMo Guardrails.
- How do I install llamaguard?
- Run /plugin marketplace add Orchestra-Research/AI-Research-SKILLs in your agent, then reload your skills. Review the source at https://github.com/Orchestra-Research/AI-Research-SKILLs before installing.
- Is llamaguard free to use?
- Yes. llamaguard is free and open source under the MIT license, so you can read, run, and adapt it within that license's terms.
- Where does llamaguard come from?
- llamaguard ships inside Orchestra-Research/AI-Research-SKILLs, a repository that contains 41 catalogued skills in total. The repository's 11,807 GitHub stars apply to that whole collection, not to this skill on its own.
Related skills
More Backend →Fast tokenizers optimized for research and production. Rust-based implementation tokenizes 1GB in <20 seconds. Supports BPE, WordPiece, and Unigram algorithms. Train custom vocabularies, track alignments, handle padding/truncation. Integrates seamlessly with transformers. Use when you need high-performance tokenization or custom tokenizer training.
Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.
High-level PyTorch framework with Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks system, and minimal boilerplate. Scales from laptop to supercomputer with same code. Use when you want clean training loops with built-in best practices.
MUST USE for ANY work on .py .pyi .rs .ts .tsx .mts .cts .go files. One philosophy: strict types, modern stacks (Pydantic v2 / serde+thiserror / Zod / gin+sqlc+pgx+slog), modern toolchains (uv+basedpyright+ruff / cargo+clippy+miri / Bun+Biome+tsc / gofumpt+golangci-lint v2+nilaway+go-race), parse-don't-validate, exhaustive match, typed errors, no any/unwrap/panic, 250 LOC ceiling, TDD, consumer-routed logging. Routes to references/{python,rust,typescript,rust-ub,go}/ + references/logging.md. Triggers: write/edit Python/Rust/TypeScript/Go code, new project, gin server, bubbletea TUI, CJK IME, connect-go RPC, sqlc pgx, branded ids, exhaustive match, unsafe Rust, miri, oversized file, refactor, TDD, e2e test, logging, log levels, structured logging, observability, arena, allocator, bumpalo, const fn, const generics, comptime, zero-alloc, bitfield, repr, scopeguard, errdefer, Zig-like, zerocopy, packed struct.