Luxas is an open-source multi-agent research system that takes a research topic through literature searches, experiments, reviews, and LaTeX manuscript compilation. It is intended for autonomous scientific research, using separate agents to read papers, run tests, create figures, and write reports. The catalogue lists skills associated with this workflow.
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 skills add Muuuun/luxas --skill qec-constructgit clone --depth 1 https://github.com/Muuuun/luxasWrote 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/skills/muuuun/luxas/qec-construct)<a href="https://agentmods.dev/skills/muuuun/luxas/qec-construct"><img src="https://agentmods.dev/badge/skills/muuuun/luxas/qec-construct/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/skills/muuuun/luxas/qec-construct"><img src="https://agentmods.dev/badge/skills/muuuun/luxas/qec-construct.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Agent Snooping · line 21 Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
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 | $0.00148 | $0.01619 |
| Opus 5.5 | $0.00059 | $0.00648 |
| Sonnet 5.5 | $0.00030 | $0.00324 |
| Haiku 4.5 | $0.00015 | $0.00162 |
Grade A, and why
qec-construct 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 1mo 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 — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.
qec-construct — a QEC verifier-in-the-loop
Routing (brain: read this section, forward the rest)
Applies when RESEARCH.md's goal is a new or improved QEC code / code family / transversal-gate construction — construct, invent, design, discover, improve distance/FOM, and search-phrased variants ("find codes beating X") equally.
Does NOT apply to magic-state cultivation, lattice surgery, decoder, or imaging projects: the gate check is abelian/CCZ-only and those are out of scope regardless of wiring.
Brain's only two obligations (do not run the loop yourself):
- Read this file BEFORE writing notes/plan.md.
- Every code-construction/code-search experiment spawn task MUST (a) include the
literal path
skills/qec-construct/SKILL.mdwith the instruction to read it in the Design phase and wrapscripts/qverifyas a tool, and (b) frame the experiment as a construct-loop (propose a parametrized ALGEBRAIC construction → qverify → debug the failure), never as random/grid sampling over supports — sampling is the mode that capped at FOM ~10.4 vs frontier 45.5.
The one rule: CONSTRUCT, don't search
Broad/random/hill-climb search over code supports is a known dead end for this problem (empirically caps far below the frontier; the good codes are rare isolated optima). Your job is to propose a parametrized ALGEBRAIC construction — a group + a generating rule for the supports + (for a new gate) the cup-product / Leibniz conditions — and debug it against the sound verifier, not to sample points and hope.
If you find yourself enumerating random supports, stop: that is the wrong mode.
The verifier: scripts/qverify (this is your Lean)
Call it at high frequency. Input a construction spec (JSON), get a SOUND verdict — including which condition failed, so you can fix it.
echo '{"family":"abelian","group_shape":[3,4,5],
"supp_a":[[0,1,3],[0,3,0],[2,1,3],[2,3,1]],
"supp_b":[[1,1,4],[1,2,2],[2,1,1],[2,2,4]],
"supp_c":[[0,0,4],[2,0,1]],
"frontier_fom":14.4}' | scripts/qverify
What ships with it
8 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- scripts/kernel/ccz_subrank.py 19 KB runs code
- scripts/kernel/cheap_screen.py 7.3 KB runs code
- scripts/kernel/group.py 5.2 KB runs code
- scripts/kernel/tricycle_code.py 5.7 KB runs code
- scripts/kernel/tricycle_distance.py 22 KB runs code
- scripts/kernel/tricycle_stcp.py 14 KB runs code
- scripts/qverify 7.3 KB
- tests/smoke_test.py 4.2 KB runs code
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.
- 1mo ago First seen · 99 lines · 148 tokens per session scan A 896ef0ce6234
qec-construct is a skill published in the GitHub repository Muuuun/luxas (1,169 stars, last pushed 5d ago), licensed MIT. It adds 148 tokens to every session and 1,619 once invoked, about $0.0006 per session on Opus 5.5. 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 skills, from other repositories
research-rigor
A structured method for doing research and analysis with explicit assumptions, evidence checks, scoring criteria, independent review, and later accuracy checks. The description is mainly in Chinese and does not define a specific software integration.
nature-academic-search
Multi-source literature search, citation verification, MeSH search strategy, citation file management (.nbib/.ris/.bib conversion), and reference management (BibTeX, related articles, ID conversion) via MCP tools (PubMed, CrossRef, arXiv). Use when the user needs coordinated multi-step literature workflows beyond a…
ml-paper-writing
Write publication-ready ML/AI papers for NeurIPS, ICML, ICLR, ACL, AAAI, COLM. Use when drafting papers from research repos, structuring arguments, verifying citations, or preparing camera-ready submissions. Includes LaTeX templates, reviewer guidelines, and citation verification workflows.
neuroskill-bci
Connect to a running NeuroSkill instance and incorporate the user's real-time cognitive and emotional state (focus, relaxation, mood, cognitive load, drowsiness, heart rate, HRV, sleep staging, and 40+ derived EXG scores) into responses. Requires a BCI wearable (Muse 2/S or OpenBCI) and the NeuroSkill desktop app…
sparse-autoencoder-training
Provides guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network activations into interpretable features. Use when discovering interpretable features, analyzing superposition, or studying monosemantic representations in language models.
distributed-llm-pretraining-torchtitan
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.