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/10cg/aria/claude-mdgit clone --depth 1 https://github.com/10CG/AriaWrote 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/10cg/aria/claude-md)<a href="https://agentmods.dev/instructions/10cg/aria/claude-md"><img src="https://agentmods.dev/badge/instructions/10cg/aria/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 | $0.04088 | $0.04088 |
| Opus 5 | $0.02044 | $0.02044 |
| Sonnet 5 | $0.00818 | $0.00818 |
| Haiku 4.5 | $0.00409 | $0.00409 |
Grade A, and why
Aria CLAUDE.md scanned grade A with 1 finding 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 3d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
forgejo <METHOD> <ENDPOINT> [curl options] How it starts
The opening of the file, as written. The whole thing — 152 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Aria - AI-DDD Methodology
项目本质: AI-DDD 方法论的定义与端到端参考实现 (v1.x 方法论 + v2.0 自主运行时) 核心假设: AI 不仅是协作者 (v1.x), 更是 SDLC 的自主执行者 (v2.0) 版本: 2.0.0
文档边界
- README.md = 人类用户 ("如何使用 Aria") | CLAUDE.md = AI 助手 ("如何理解 Aria 项目")
- CLAUDE.md 卫生 (Option A,
standards/conventions/claude-md-hygiene.md, 对齐 Claude Code 官方 memory 指南 ≤200 行目标): 只放稳定的「每 session 都该知道的事实」。不放: 版本 changelog (→aria/CHANGELOG.mdSOT) / session 进展流水 (→docs/handoff/, Rule #9) / Skill 设计内部术语 (→ 各 SKILL.md 按需加载; 抄进本文件会污染 AB baseline, 实证 aria-plugin #116)。「项目状态」段 = live 覆写, 预算 15-20 行。enforcement: state-checkclaude-md-changelog-free。
工作语言
中文叙述为主。保留英文技术 token: 代码 / 命令 / 路径 / SHA / branch / PR# / 版本号 / spec 术语 / Skill·Agent 名 / memory 引用。叙述 / 解释 / 建议 / 询问 (含 AskUserQuestion options) / handoff prose 用中文。详: memory user_chinese_conversation_default。
项目定位
方法论研究项目 (非框架实现): 探索 AI Agent 深度参与软件工程全流程。
- 协作模式: AI 理解 → 人类确认 → 协作交付 (对比传统: 人类主导 → AI 辅助)
- v1.x = 人类 + Claude Code (interactive); v2.0 = 同一套方法论放到无人值守 (Hermes Layer 1 + aria-runner Layer 2)
- 研究目标: 可重现的 AI 协作流程 / 最小化上下文传递成本 / 结构化决策记录
- 运行时架构详见 aria-orchestrator/docs/architecture-decisions.md (AD1-AD12)
核心概念
十步循环: A 规划 (A.0 状态扫描 → A.1 规范创建 → A.2 任务规划 → A.3 Agent 分配) → B 开发 (B.1 分支创建 → B.2 执行验证; Skill 变更时含 benchmark) → C 集成 (C.1 提交 → C.2 合并) → D 收尾 (D.1 进度更新 → D.2 归档)。SOT: standards/core/ten-step-cycle/。
两层 AI 分工 (v2.0): Layer 1 主管 (Hermes + Luxeno-routed GLM, PM 角色: triage/派发/审批; 只加载 ~1K token 元知识, 不加载 aria-plugin, AD7) / Layer 2 工程师 (aria-runner 容器 + Claude Code + aria-plugin 完整加载, 执行完整十步循环)。两层用拟人命令 (自然语言 YAML) 通信, 非结构化 RPC (AD1 + AD6)。
OpenSpec 需求规范: Level 1 = Skip (简单修复) / Level 2 = proposal.md / Level 3 = proposal.md + tasks.md。
协作原则: 规范先行 (先 spec 后代码) / 小步迭代 (任务 4-8h 粒度) / 文档同步 (文档过时 = AI 误解) / 向后兼容 (破坏性变更须 MAJOR)。
信息地图
| 子模块/目录 | 职责 |
|---|---|
standards/ |
方法论定义 (十步循环 / OpenSpec / conventions) |
aria/ |
工具集 Plugin (Skills + Agents + Hooks) |
aria-plugin-benchmarks/ |
Skill AB 基准测试 (固定套件 / 结果存档 / 运维手册) |
docs/handoff/ |
Session handoff records (Rule #9 canonical) |
aria-orchestrator/ |
v2.0 运行时 Layer 1/2 (仅 10CG Lab 内部) |
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.
- 3d ago First seen · 152 lines · 4,088 tokens per session scan A a6b102152b7f
Aria CLAUDE.md is an instructions file published in the GitHub repository 10CG/Aria (2 stars, last pushed 4d ago), licensed MIT. It adds 4,088 tokens to every session, about $0.0204 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
appointme CLAUDE.md
Instructions for bravodev-hub/appointme, covering claude.md, build and development commands, running the full stack (recommended), backend only and frontend only.
Praxis CLAUDE.md
Instructions for BackToCimaCoppi/Praxis, covering praxis 贡献规范（ai 与人共同遵守）, 1. skill 格式, 2. 文档格式, 3. 脱敏红线（开源仓硬约束，最高优先级） and 4. 新增 skill 自检清单（提 pr 前逐条打勾）.
add AGENTS.md
Instructions for MountainUnicorn/add, covering add, engagement protocol, spec-first invariants, maturity & autonomy ceiling and currently active spec.
ai-interview-kit CLAUDE.md
Instructions for CyannSHI/ai-interview-kit: This project includes 3 AI skills for end-to-end AI phone interview support. Each skill asks the user to choose their language (English/中文) at the start, then adapts accordingly.
SDLC-Enterprise-Framework CLAUDE.md
Claude Code instructions for Minh-Tam-Solution/SDLC-Enterprise-Framework, covering claude.md - ai assistant guidelines for sdlc 6.5.0, repository purpose, key framework concepts, code folder organization (not stage-mapped) and code file naming standards (new in 4.9.1).
ai-collab-playbook AGENTS.md
AGENTS.md instructions for cnfjlhj/ai-collab-playbook, covering codex 全局指导原则, 0. 工作模式, 0.1 一路畅行模式, 1. skills and 2. 核心行为准则.