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/cybertronai/sutroyaro/gemini-mdgit clone --depth 1 https://github.com/cybertronai/SutroYaroWrote 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/cybertronai/sutroyaro/gemini-md)<a href="https://agentmods.dev/instructions/cybertronai/sutroyaro/gemini-md"><img src="https://agentmods.dev/badge/instructions/cybertronai/sutroyaro/gemini-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 5.1 | $0.02086 | $0.02086 |
| Opus 5 | $0.01043 | $0.01043 |
| Sonnet 5 | $0.00417 | $0.00417 |
| Haiku 4.5 | $0.00209 | $0.00209 |
Grade A, and why
SutroYaro GEMINI.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 5d 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 — 153 lines — stays where its author put it; the contents beside it link to each section on GitHub.
GEMINI.md - Sutro Group Research Workspace
Project Context
This is a research workspace for the Sutro Group, a study group exploring energy-efficient AI training. The group meets weekly at South Park Commons in San Francisco.
Read These First
- LAB.md — Protocol for running experiments (templates, lifecycle, rules)
- AGENT.md — Machine-executable experiment loop for autonomous sessions
- DISCOVERIES.md — What's proven so far (read before every experiment)
- CONTRIBUTING.md — How external contributors submit experiments and findings
- TODO.md — Open research tasks
- docs/tasks/INDEX.md — Current task tracker with priorities
- docs/research/survey.md — Practitioner's Field Guide ranking all 36 experiments
- docs/research/peer-research-protocol.md — Full design doc for multi-researcher autonomous research
Core Concepts
- Sparse Parity: The benchmark task — learn XOR/parity from random {-1,+1} inputs. n=20 bits, k=3 secret, 17 noise. The "drosophila" of energy-efficient training.
- Average Reuse Distance (ARD): Proxy metric for energy efficiency. Small ARD = data stays in cache = cheap. Large ARD = expensive external memory access.
- Data Movement Complexity (DMC): Better proxy metric (Ding et al., arXiv:2312.14441). DMC = sum of sqrt(stack_distance) for all float accesses. Tracks alongside ARD in MemTracker. Baseline: ARD 4,104 / DMC 300,298.
- Cache Energy Model: register 5pJ, L1 (64KB) 20pJ, L2 (256KB) 100pJ, HBM 640pJ per float access (Bill Dally numbers).
- TrackedArray / Auto DMD:
TrackedArraywraps numpy arrays so every operation (ufuncs, indexing, slicing) auto-records reads and writes on anLRUStackTracker. Removes manual instrumentation errors. Seedocs/research/tracked-numpy.md.
Current Best Methods
| Method | Time (n=20/k=3) | ARD | DMC | Notes |
|---|---|---|---|---|
| KM-min (1 sample) | ~0.001s | 20 | 3,578 | New DMC leader. 1 influence sample suffices for parity. |
| GF(2) Gaussian Elimination | 509 us | ~420 | ~203K | 240x faster than SGD, k-independent. Auto-tracked via TrackedArray; old harness reported 8,607. |
| KM Influence Estimation | 0.001-0.006s | 92 | 20,633 | ARD leader. 5 influence samples per bit. |
| SMT Backtracking | 0.002s | 3,360 | 348,336 | Constraint satisfaction approach |
| SGD (baseline) | 0.12s | 8,504 | 1,278,460 | LR=0.1, batch=32, hidden=200 |
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.
- 5d ago First seen · 153 lines · 2,086 tokens per session scan A 688d569116e8
SutroYaro GEMINI.md is an instructions file published in the GitHub repository cybertronai/SutroYaro (16 stars, last pushed 3mo ago), licensed Unlicense. It adds 2,086 tokens to every session, about $0.0104 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
PINA AGENTS.md
AGENTS.md instructions for PINA-org/PINA, covering pina — physics-informed neural architectures, quick reference, workflow: problem → model → solver → trainer, problem types and condition types.
ClimaAtmos.jl AGENTS.md
AGENTS.md instructions for CliMA/ClimaAtmos.jl, covering climaatmos agent guide, ecosystem guidelines, repo-specific guidelines, local norms and self-correction.
tensorcircuit-ng AGENTS.md
AGENTS.md instructions for tensorcircuit/tensorcircuit-ng, covering tensorcircuit-ng repository guide for ai agents, mission, non-negotiable rules, environment rules and where to look first.
tpu_performance_autoresearch_wiki GEMINI.md
Gemini CLI instructions for vlasenkoalexey/tpu_performance_autoresearch_wiki, covering gemini/antigravity operating rules, platform adaptation (claude code → gemini/antigravity), 1. skills — native, no emulation, 2. never-stop hook & retrospectives and 3. session and transcript resolution.
tpu_performance_autoresearch_wiki CLAUDE.md
Claude Code instructions for vlasenkoalexey/tpu_performance_autoresearch_wiki, a project described as: Wiki built using Karpathy method containing information about TPU performance optimizations and hooking it up to autoresearch optimization engine.
co-mathematician CLAUDE.md
Instructions for VeryMath/co-mathematician, covering claude.md, repository contract, required flow, claude code operating notes and harness commands.