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 instructions/smsharma/clax/claude-mdgit clone --depth 1 https://github.com/smsharma/claxWrote 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/smsharma/clax/claude-md)<a href="https://agentmods.dev/instructions/smsharma/clax/claude-md"><img src="https://agentmods.dev/badge/instructions/smsharma/clax/claude-md.svg" alt="Measured on agentmods" 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 | $0.05537 | $0.05537 |
| Opus 5 | $0.02769 | $0.02769 |
| Sonnet 5 | $0.01107 | $0.01107 |
| Haiku 4.5 | $0.00554 | $0.00554 |
Grade A, and why
clax 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 yesterday.
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 — 522 lines — stays where its author put it; the contents beside it link to each section on GitHub.
clax Development Guide
CRITICAL: HPC environment rules
Filesystem
- NEVER use
find /,find /ocean, or scan outside the project directory. The HPC filesystem has millions of files and these commands will hang forever. - Reference CLASS source:
../class_public-3.3.4/— ALWAYS use this path. Key files:source/perturbations.c,source/transfer.c,source/harmonic.c,source/thermodynamics.c,source/background.c - GPU diagnostic scripts:
diags/— reusable standalone scripts for targeted accuracy investigation (C_l convergence, perturbation var comparison, x_e oracle injection, source decomposition, timing). Seediags/README.md. - Project root: the current working directory (use
.or relative paths) - Only search within
.and../class_public-3.3.4/ - Use
greporGrepto search file contents — neverfindwith broad paths.
GPU access
You are running directly on a GPU compute node (V100-32GB). Run python and pytest directly — no wrapper needed:
python scripts/gpu_planck_test.py
pytest tests/ --fast -x -q
Do NOT use gpu-run.sh — that was for login-node use only.
What is this?
A fully differentiable reimplementation of the CLASS Boltzmann solver in JAX. Full architecture and physics details are in DESIGN.md. This file covers development instructions and conventions.
Quick reference
- Reference CLASS:
../class_public-3.3.4/ - Design document:
DESIGN.md(read this first) - Progress log:
CHANGELOG.md
Current performance (Mar 14, 2026)
Two presets for different use cases:
| Preset | Total (cached) | GPU | Accuracy | Use case |
|---|---|---|---|---|
| fit_cl | 34s | V100 | TT/EE < 1.5% at l≤500 | HMC / fitting |
| planck_cl | ~487s | H100 | TT/EE < 0.2% at l≤2000 | Science-grade |
fit_cl uses 20 k/decade, l_max=17, table-based Bessel (harmonic in 2.4s), ncdm_q_size=0, rtol=1e-3, max_steps=1024. Perturbation ODE at ~30s is the floor.
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.
- yesterday Changed · +28 lines · +427 tokens per session 8097be7bde5e
- 5d ago First seen · 494 lines · 5,110 tokens per session scan A 1d79be5c387d
clax CLAUDE.md is an instructions file published in the GitHub repository smsharma/clax (60 stars, last pushed today), licensed MIT. It adds 5,537 tokens to every session, about $0.0277 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 2026-08-30.
Other instructions, from other repositories
bio-gene-to-reference-tree copilot-instructions.md
Copilot instructions for Hongda-Zhao/bio-gene-to-reference-tree, a project described as: Auditable agent skill for resolving protein queries, selecting references, and planning reproducible phylogenetic trees.
braina GEMINI.md
Instructions for brainets/braina, covering project: braina (brain interaction analysis), 1. project context & purpose, 2. commands, verify environment (all core dependencies) and run the verification test suite for frites + hoi.
research-automation CLAUDE.md
Instructions for lucafusarbassini/research-automation, covering ricet - research automation framework, project overview, claude-flow mcp, workflow habits and file organization.
shannon-prover CLAUDE.md
Claude Code instructions for SkyShannonProver/shannon-prover, covering shannon prover: claude entry point, current boundary, easycrypt environment, eval safety and current documentation.
SciCrucible CLAUDE.md
Claude Code instructions for Xinyang-Li666/SciCrucible, covering 科学知识库, 知识库结构, 可用命令, 项目目录 and 工作原则.
torchcell CLAUDE.md
Claude Code instructions for Mjvolk3/torchcell, covering vision -- the virtual cell (north star), for local configs, git worktrees, programming guide and provenance & reproducibility.