Borrowing it
Nothing to install: this file belongs to l-nmch/hailo-10h-llm-compiler. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/l-nmch/hailo-10h-llm-compiler/main/CLAUDE.mdgit clone --depth 1 https://github.com/l-nmch/hailo-10h-llm-compilerWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/l-nmch/hailo-10h-llm-compiler/claude-md)<a href="https://agentmods.dev/instructions/l-nmch/hailo-10h-llm-compiler/claude-md"><img src="https://agentmods.dev/badge/instructions/l-nmch/hailo-10h-llm-compiler/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/l-nmch/hailo-10h-llm-compiler/claude-md"><img src="https://agentmods.dev/badge/instructions/l-nmch/hailo-10h-llm-compiler/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00896 | $0.00896 |
| Opus 5 | $0.00448 | $0.00448 |
| Sonnet 5 | $0.00179 | $0.00179 |
| Haiku 4.5 | $0.00090 | $0.00090 |
Grade A, and why
hailo-10h-llm-compiler CLAUDE.md scanned grade A with 0 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 11d 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 66 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Working in this repository
This file is for any agent (human or AI) picking up the investigation. Read docs/status.md first for the current state, then this page for how to work without breaking the project's invariants.
Non-negotiable ground rules
- The chip runs AS-IS. Read/decompile/probe freely; never patch or flash a modified firmware or driver image, even reversibly. The entire point of this project is compiling a self-made model for the stock device.
- Nothing installed on the host. All tooling — DFC, PyTorch, HailoRT
clients — runs inside the Docker images in docker/. Don't
pip installorapt-geton bare metal. - One command, one container. Each independent DFC/PyTorch invocation
gets its own
docker run --name <x>; don'tdocker execinto a shared long-lived container — state becomes unreadable and non-reproducible. Read results back viadocker logs. - Proprietary material policy — see CONTRIBUTING.md.
No DFC wheels, official HEFs/
.alls/.hn, firmware images, or long verbatim excerpts of proprietary source, ever, in any commit. - No dates or timelines anywhere in committed content — keeps the repo evergreen and diff-stable (CONTRIBUTING.md).
Working patterns that avoid known traps
- One persistent session for serving, subprocess-per-attempt for
unattended/CI generation — see
Runtime-Pitfalls for why (interrupt/ioctl
failure mode). Never open a fresh VDevice/LLM per attempt in a serving
loop; use
clear_context(). - GPU quantization: try plain GPU before
use_saitama=True. The Saitama PyTorch engine exists to unlock AMD/ROCm GPUs (the stock optimizer is CUDA/TensorFlow-bound) — see Quantization-Deep-Dive. Don't reach for it on NVIDIA where the default path already works. - Numerical comparisons are one batched
infer()call, teacher-forced,batch_size == calibset_size. Never loop token-by-token for a cosine check — it's slow and leaks memory for no accuracy gain. - Memory-constrained hosts (small SBCs, or a 32 GB workstation running
DFC's optimizer): never load a multi-GB file whole into RAM, and never
run an uncapped
optimize()— see the memory-discipline note in Runtime-Pitfalls. - Degraded/unreachable device state: recover with a full power cycle of the host only. Do not loop software rescans, power-state pokes, or driver unbind/rebind over a remote link — this has wedged hosts harder every time it was tried.
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.
- 11d ago First seen · 66 lines · 896 tokens per session scan A cc65e43055de
hailo-10h-llm-compiler CLAUDE.md is an instructions file published in the GitHub repository l-nmch/hailo-10h-llm-compiler (14 stars, last pushed 10d ago), licensed MIT. It adds 896 tokens to every session, about $0.0045 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).