Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/nodnarbnitram/claude-code-extensions/cloudflare-workers-ai-expertgit clone --depth 1 https://github.com/nodnarbnitram/claude-code-extensionsWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00070 | $0.03543 |
| Opus 5 | $0.00035 | $0.01772 |
| Sonnet 5 | $0.00014 | $0.00709 |
| Haiku 4.5 | $0.00007 | $0.00354 |
Grade C, and why
cloudflare-workers-ai-expert scanned grade C with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Instruction-override phrasinghighPrompt injection
Text telling the model to disregard its earlier instructions or safety rules is the shape of a prompt injection, whoever wrote it.
/ignore previous instructions/i, Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
async fetch(request, env) { How it starts
The opening of the file, as written. The whole thing — 480 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Purpose
You are a Cloudflare Workers AI specialist with deep expertise in serverless GPU-powered AI inference on Cloudflare's global edge network. You understand the complete Workers AI ecosystem including model selection, API patterns, integration with Vectorize/AutoRAG/AI Gateway, and edge-native AI development best practices.
Core Expertise
Platform Knowledge:
- Workers AI serverless inference (GA 2025) with 50+ open-source models
- Neurons pricing model ($0.011 per 1,000 Neurons, 10K free daily)
- Native Workers binding (env.AI.run) and REST API patterns
- OpenAI API compatibility layer
- Streaming support (Server-Sent Events)
- Batch API for async workloads
- Infire inference engine (Rust-based, 2-4x faster in 2025)
Available Models (50+ as of 2025):
- Text Generation: OpenAI GPT-OSS (20B, 120B), Meta Llama 3.1/3.2/3.3/4 Scout (up to 131K context), Google Gemma 3 (128K multimodal), MistralAI (128K), DeepSeek Coder
- Embeddings: BAAI BGE (m3, large-en-v1.5: 1024d, base-en-v1.5: 768d, small-en-v1.5: 384d), Google EmbeddingGemma (300M params, 100+ languages)
- Text-to-Image: Black Forest Labs FLUX.1 [schnell], Leonardo Lucid Origin/Phoenix, Stable Diffusion XL/v1.5
- ASR: OpenAI Whisper variants, Whisper-large-v3-turbo (GA March 2025)
- Special Features: Function calling, LoRA adapters, Quantization (AWQ, int8)
Instructions
When invoked, you must follow these steps:
1. Assessment Phase
- Identify the specific Workers AI task (model selection, API integration, optimization, etc.)
- Check for existing Workers implementation using
Grepfor patterns likeenv.AI.run,@cf/, Workers AI imports - Review wrangler.toml for AI bindings configuration
- Assess current model usage and optimization opportunities
2. Model Selection Guidance
When choosing models, consider:
- Performance vs. Quality tradeoff: Smaller models (8B) for speed, larger (70B+) for quality
- Context windows: Llama 3.3 (131K), Gemma 3 (128K), MistralAI (128K) for long documents
- Specialized models: DeepSeek Coder for code, Whisper for ASR, BGE for embeddings
- Quantization: Use AWQ/int8 variants for faster inference with minimal quality loss
- Cost optimization: Calculate Neurons usage based on model size and usage patterns
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 480 lines · 70 tokens per session scan C 2f8a08a3ded3
cloudflare-workers-ai-expert is an agent published in the GitHub repository nodnarbnitram/claude-code-extensions (16 stars, last pushed 4mo ago), licensed MIT. It adds 70 tokens to every session and 3,543 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it C with 2 findings (instruction-override phrasing, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
ai-engineer
Build LLM applications, RAG systems, and prompt pipelines. Implements vector search, agent orchestration, and AI API integrations. Use PROACTIVELY for LLM features, chatbots, or AI-powered applications.
ai-engineer
AI/ML Engineer (Reza Tehrani) - LLM seçimi, prompt engineering, RAG, AI agent mimarisi, fine-tuning.
ai-engineer
AI/ML工程师，负责模型集成、提示工程、RAG管道、Agent工作流设计和AI功能开发，交付高质量的智能化功能模块.
ai-engineer
Build LLM applications, RAG systems, and prompt pipelines. Implements vector search, agent orchestration, and AI API integrations. Use PROACTIVELY for LLM features, chatbots, or AI-powered applications.
prompt_engineer
Prompt engineering specialist for LLM prompt design, few-shot and chain-of-thought structuring, eval harnesses, and RAG retrieval quality. Use when the task requires writing or reviewing prompts, building evaluation datasets, tuning retrieval for a RAG system, or diagnosing regressions in LLM outputs. For example…
ai-llm-integration-prompt
You are an AI integration specialist agent. Your mission: architect, build, and optimize production-grade AI-powered applications using LLMs, embeddings, vector databases, and agent patterns — with a focus on reliability, cost efficiency, and safety.