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/breakstring/agent-deck/agents-mdgit clone --depth 1 https://github.com/breakstring/agent-deckWhat 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 | $0.02638 | $0.02638 |
| Opus 5 | $0.01319 | $0.01319 |
| Sonnet 5 | $0.00528 | $0.00528 |
| Haiku 4.5 | $0.00264 | $0.00264 |
Grade A, and why
agent-deck AGENTS.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 2d 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 — 163 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
本文件是 Agent Deck 项目的项目级 Coding Agent 指南。它补充全局规则，优先用于本仓库 /Users/kenn/Projects/agent-deck 内的开发、审查、测试和文档维护。
项目定位
Agent Deck 是一个本机 AI Agent 硬件控制台桥接项目。第一阶段目标是打通 macOS + 妙联宝 N4 Pro + Codex 的最快闭环：
- 采集 Codex 的运行状态、工具调用、审批请求和会话事件。
- 将状态归约成统一
AgentState，再生成硬件无关的LayoutPlan。 - 在 N4 Pro 或 fake hardware surface 上展示 Agent 状态。
- 通过硬件输入选择 Agent、聚焦上下文，并在可控范围内响应 Codex
PermissionRequest。
长期方向是支持更多 AI Agents 和更多妙联宝硬件。不要把实现写死成“Codex 专用按钮脚本”或“N4 Pro 专用渲染脚本”。项目核心边界应该保持为：
Agent ingress -> NormalizedEvent -> AgentStateStore -> DeckMode/LayoutPlan -> HardwareSurface -> InteractionIntent/ActionExecutor
必读文档
开始非平凡开发前，先按任务范围阅读以下文档：
docs/superpowers/specs/2026-06-12-agent-deck-analysis.md：总体需求、长期抽象、安全边界和非目标。docs/superpowers/specs/2026-06-12-agent-deck-mvp-design.md：第一版macOS + N4 Pro + Codex的 MVP 设计。docs/references/agent-deck-roadmap.md：长期 roadmap，后续任务拆分优先参考这里。docs/references/stream-dock-scenes-research-2026-06-12.md：妙联宝官方场景概念和 Agent Deck 内部 DeckMode 的边界。docs/references/gemini-agent-deck-analysis-2026-06-12.txt：Gemini 早期分析归档，只能作为参考，不可直接照搬其中伪代码。
如果需求与 roadmap 冲突，以用户最新要求为准，并同步更新 roadmap 或在交付说明中指出文档待更新。
工作流
- 默认中文回复，文档、注释、commit 信息也使用中文，除非用户明确要求英文。
- 修改前先看
git status --short --branch，确认是否有用户未提交改动。不要回退、覆盖或清理不是自己产生的改动。 - 需要隔离开发时使用 git worktree，目录放在
.worktrees/下。该目录必须保持被.gitignore忽略。 - Python 依赖、命令和测试统一通过
uv执行。不要手工维护裸pip流程作为项目默认路径。 - 常规子代理默认使用
gpt-5.5，推理强度medium，除非用户明确指定。 - 交付前运行与改动范围匹配的测试。能跑全量时优先执行
uv run pytest -q。 - 不做破坏性 git 操作，例如
git reset --hard、git checkout -- <file>，除非用户明确要求。
当前代码结构
核心模块职责如下：
src/agent_deck/core/events.py：统一事件模型、payload 脱敏、递归冻结、timezone-aware 时间校验。src/agent_deck/core/state.py：事件到 Agent 状态的内存归约。src/agent_deck/core/decisions.py：审批决策 broker、超时、默认 deny、跨线程/跨 event loop 等待。src/agent_deck/core/modes.py：DeckMode 和当前选择状态。src/agent_deck/rendering/layout.py：由状态和决策生成硬件无关布局计划。src/agent_deck/hardware/fake.py：无真实 I/O 的 fake hardware surface，用于测试和 daemon MVP。src/agent_deck/hardware/streamdock_probe.py：真实 StreamDock 设备只读诊断探针。src/agent_deck/server/app.py：本地 FastAPI daemon API，当前使用内存 runtime 和 fake hardware。src/agent_deck/cli.py：agent-deckd、agent-deckctl、agent-deck-codex-hook三个 CLI 入口。tests/：当前核心行为的自动化测试。
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.
- 2d ago First seen · 163 lines · 2,638 tokens per session scan A 7cb876649f2d
agent-deck AGENTS.md is an instructions file published in the GitHub repository breakstring/agent-deck (38 stars, last pushed 4d ago), licensed MIT. It adds 2,638 tokens to every session, about $0.0132 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
sandvault AGENTS.md
Instructions for webcoyote/sandvault, covering agents.md and testing.
codex-project-skill-manager AGENTS.md
Instructions for WildPigKing/codex-project-skill-manager, covering codex project & skill manager, product boundary, startup routing, consent contract and repository records.
codex-token-meter AGENTS.md
Instructions for prefect12/codex-token-meter, covering ai token meter development guide, project shape, build and verification, git workflow and ai token meter test loop.
duetlens CLAUDE.md
Claude Code instructions for xieziyu/duetlens, covering duetlens, 文档约定, 版本控制, 语言 and 文件命名.
codex-switch AGENTS.md
AGENTS.md instructions for ChArLiEdance/codex-switch, covering repository agents and platform isolation.
speech-swift AGENTS.md
Instructions for soniqo/speech-swift, covering agent instructions, communication style, workflow, running tests and benchmarks — sequential and memory-aware and git conventions.