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/notatestuser/ds4-control/claude-mdgit clone --depth 1 https://github.com/notatestuser/ds4-controlWrote 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/notatestuser/ds4-control/claude-md)<a href="https://agentmods.dev/instructions/notatestuser/ds4-control/claude-md"><img src="https://agentmods.dev/badge/instructions/notatestuser/ds4-control/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 | $0.01320 | $0.01320 |
| Opus 5 | $0.00660 | $0.00660 |
| Sonnet 5 | $0.00264 | $0.00264 |
| Haiku 4.5 | $0.00132 | $0.00132 |
Grade A, and why
ds4-control 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 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 — 71 lines — stays where its author put it; the contents beside it link to each section on GitHub.
DS4 Control — project guide for Claude
Purpose
A macOS menu-bar control pane for DeepSeek V4 running locally on Apple Silicon via
antirez/ds4. It launches, supervises, and monitors a local
ds4-server child process; lets you pick V4 Pro or V4 Flash (0731 weights); downloads GGUF weights;
shows live unified-memory / GPU / CPU / power widgets; provides a built-in chat; and can open a
coding agent (pi or claude) in Terminal pointed at the local server.
It does no embedded inference — all inference is delegated to ds4-server. This app only
supervises that process and surfaces system metrics + a chat/agent front end.
DeepSeek V4 preview maintenance
The current 0731 DeepSeek V4 GGUFs are preview artifacts, not the final GA release. When V4
leaves preview, update the pinned GGUF names/URLs and external/ds4 revision together. Re-read
the GA files' exact byte sizes into Quant.ggufBytes, verify the Metal context-allocation formula,
shared graph-workspace formula, and persistent backend-scratch bounds against that ds4 revision,
and refresh the feasibility/variant tests and documented memory tiers.
Do not carry the preview size constants or allocator assumptions into a GA release unchanged.
Stack
- SwiftPM executable target
DS4Control(Swift 6 mode, macOS 14+).swift-tools-version: 6.3. - SwiftUI menu-bar app (
MenuBarExtra, LSUIElement/.accessory— no dock icon or window until you click the menu-bar item). LinksIOKit+ privateIOReportfor power/frequency metrics. ds4lives as a git submodule atexternal/ds4(provides theds4-serverbinary).
Dev workflow
From the repo root (/Users/luke/dev26/ds4_workspace/ds4-control):
swift build # build
swift test # run all tests (authoritative — trust the compiler over SourceKit squiggles)
DS4_DIR="$PWD/external/ds4" .build/debug/DS4Control # run the dev app
DS4_DIR points the app at the bundled submodule so Start can spawn ds4-server. It's a menu-bar
app — after launch, find its icon in the macOS menu bar; the popup's gear/chat/terminal icons open
Settings, the chat window, and the agent launcher.
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 · 71 lines · 1,320 tokens per session scan A 7e27cae36747
ds4-control CLAUDE.md is an instructions file published in the GitHub repository notatestuser/ds4-control (810 stars, last pushed 19d ago), licensed MIT. It adds 1,320 tokens to every session, about $0.0066 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
DeepSeek-Harness-Desktop AGENTS.md
Instructions for web-casa/DeepSeek-Harness-Desktop, covering agents.md — dsh desktop, 项目本质, 不可违背的安全不变量, 版本单一事实源（改动版本时必须同时改全部） and harness 升级启动契约（bump pin 前必读）.
carboncode AGENTS.md
Instructions for Yapie0/carboncode, covering agents.md, project context, current status, product direction and chosen base.
carboncode CLAUDE.md
Instructions for Yapie0/carboncode, covering claude.md, canonical rules (authoritative — follow these), dev workflow, architecture map (src/) and conventions.
DeepCopilot copilot-instructions.md
Instructions for deep-copilot/DeepCopilot, covering deepcopilot 项目级 copilot / ai 审核指引, 项目概述, 技术栈, 审查时必须关注的红线 and 代码风格.
DeepSeekMeter AGENTS.md
Instructions for CWNU-Open-Source-Community/DeepSeekMeter, covering agents.md — deepseekmeter 仓库操作规范, 1. 项目是什么, 2. 常用命令（改动后必须本地验证）, ios 版验证（核心包自测 + 工程结构静态校验必跑；有 xcode 时还会构建 app 冒烟） and android 核心单测（需 jdk 17 + android sdk；jvm 直跑无需设备）.
claude-science-api-bridge AGENTS.md
Instructions for Jyx0208/claude-science-api-bridge, covering agent operating manual, prime directive, goal, repository map and success criteria.