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/unknowbug/anchorlaw/agents-mdgit clone --depth 1 https://github.com/unknowbug/anchorlawWhat 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 | $0.02141 | $0.02141 |
| Opus 5 | $0.01071 | $0.01071 |
| Sonnet 5 | $0.00428 | $0.00428 |
| Haiku 4.5 | $0.00214 | $0.00214 |
Grade A, and why
anchorlaw 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 — 74 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Anchorlaw 项目 AGENTS.md（项目级常驻指令）
本文件是索引——协议知识按需加载（协议 v0.19 §14 Anchor Skill Manifest + §15 Execution Topology），铁律正文在对应 skill 里，不在本文件常驻。 宿主说明：DSH（DeepSeek Harness）适配层入口见
dsh/AGENTS.md（当前维护）；Reasonix 宿主格式自 v0.18 起停止维护并归档（archive/reasonix/，Fork 可恢复自行迭代）。
〇、开始工作前（每个 session 必做）
git status确认工作区状态（远程 =github.com/unknowbug/anchorlaw）- 跑测试确认基线全绿：
python -m pytest --rootdir=python python/tests -q（命令行被沙箱拦截时改用 DSH 沙箱感知包装：python dsh/scripts/run_tests_sandbox.py --rootdir=python python/tests -q --basetemp=<非点前缀路径>——DSH Windows 沙箱封存 0o700 目录导致 pytest tmp 机制失效，该脚本改 0o755 后跑基线测试；basetemp 用非点前缀目录避免 scanner 跳过隐藏目录） - 若改动涉及协议文档：同步核对 §8 Maturity 与 §11 全称声称审计表，证据必须跟着走
一、项目定位（一句话）
Anchorlaw 是语言无关的代码验证协议——唯物实践论三律的代码化：
- 第一律（可检验）：
@anchor.test把声称绑到可执行验证 - 第二律（可证伪）：
@anchor.idk诚实声明边界 + staleness（90 天自动升级） - 第三律（可挑战）：协议 §12 规则挑战流程——FP 证据强制规则降级/删除
二、主工作流（Judge 驱动流水线，v0.11）
编程是构建型任务（域内：输入契约已确认；探索型任务如 RE 在域外，见协议 §15.1/§16.1 input-contract confirmation criterion）——由 Judge 驱动：主会话扮演 Judge 角色持有验收判据，派 Scout 起草规范、派 Worker 实施，每段以「Judge 点头」收敛。需求发掘不在本协议内——它属于独立的需求协议（Scout 驱动 + 人机对话），其产出（已确认需求文档 + 技术约束规范；架构设计归 stage-1 产出）作为本流水线的输入契约。权威正文在 协议 §15（输入契约 + 四段流水线 + 角色定义），本文件只索引不复制。
输入契约 + 四段流水线（每段以 Judge 点头终止）： 0. 输入契约 — 已确认需求文档 + 技术约束规范由外部需求协议交接（主机交付即实施授权，§16.1）；Judge 从输入推导验收判据，无输入不得开工
- 实施规范 — Judge 派 scout 起草变量名/模块化/框架边界 → Judge 审过才进规划
- 实施计划 — Judge 基于规范划模块
- 并行实施 — 多 worker 按模块并行写码 → 每模块 Judge 审查，点头并入 / 打回修改
- 交付 — Judge 总审 + 隔离 judge 独立验收 → 人类授予
confirmed
验收判据先行（v0.11；v0.14 输入契约分层）：判据（声称 + @anchor.test 验证载体 + @anchor.idk 边界）由 Judge 从输入契约（外部需求 + 技术约束规范；架构设计属 stage-1 产出）推导，实施前确定；实现朝判据收敛，review 意见对照判据。权威正文 协议 §15.4。
judge 触发点与验证终止门禁（v0.12；v0.15 修订）——只索引不复制：
- Judge 五触发点（输入契约接受 / 规范审查 / 计划批准 / 模块并入 / 交付验收）+ 意见分级 → 执行清单见对应 judge skill
- 终止门禁（外部测试集 / 三层意见分级 / 判据满足即完成 + 同一判据 3 轮不收敛 → 流程硬停止 + Judge 提交 Review 情况与问题详细报告交人类判定 / §12 通道）→ 协议 §15.4；AGENTS 不复制正文——镜像不同步正是 review 循环的根因
- 关键节点隔离验收：模块并入前 + 交付前 MUST 派隔离 judge subprocess（自评≠审查）
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 · 74 lines · 2,141 tokens per session scan A db04de9d55ff
anchorlaw AGENTS.md is an instructions file published in the GitHub repository unknowbug/anchorlaw (5 stars, last pushed 6d ago), licensed MIT. It adds 2,141 tokens to every session, about $0.0107 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 2026-08-31.
Other instructions, from other repositories
agentready AGENTS.md
Instructions for ambient-code/agentready, covering agentready, commands, core, development and data flow.
agentready CLAUDE.md
Instructions for ambient-code/agentready, a project described as: Repo Optimizer: Assess git repositories for AI-assisted development readiness. Submit your score!
claude-code-templates CLAUDE.md
Instructions for Justdvp/claude-code-templates, covering claude.md, project overview, development commands, package management and application commands.
devkit CLAUDE.md
Instructions for 5uck1ess/devkit, covering claude.md, layout, architectural invariants, conventions and fixing this file.
avvarre CLAUDE.md
Instructions for PralhadYadawad/avvarre, covering avvarre — claude code instructions, what is avvarre?, available slash commands, mcp tools and project memory (.avvarre/).
claudemd CLAUDE.md
Instructions for sdsrss/claudemd, covering code graph (repo-wide ast index) and claude-mem-lite — persistent memory.