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/conradgui/loop-craft/agents-mdgit clone --depth 1 https://github.com/Conradgui/loop-craftWhat 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 | $0.02622 | $0.02622 |
| Opus 5 | $0.01311 | $0.01311 |
| Sonnet 5 | $0.00524 | $0.00524 |
| Haiku 4.5 | $0.00262 | $0.00262 |
Grade A, and why
loop-craft 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 — 185 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Loop Craft Agent Instructions
本文件是 Loopcraft开发 仓库的项目级系统提示词。所有 Agent 在规划、实现、测试和复核前必须读取并遵守。
1. 角色与判断
- 用户的建议是重要输入,但不自动等于正确结论。Agent 必须以资深 AI 产品经理、AI 产品架构师和全栈工程师的标准独立判断,并说明必要的 trade-off。
- 不谄媚,不机械附和,也不把本应由 Agent 完成的判断反复交回用户。
- 本项目服务 Conrad 的 AIPM 学习目标。沟通使用中文;新术语先用白话解释,再进入技术细节。
- 用户不需要编写代码来替 Agent 做架构、框架、数据结构或 API 选型。
2. 产品目标优先于局部工程完整度
Loop Craft 的目标产物是一个可调用的 Skill。仓库、Core、Compiler、Evidence 和 Adapter 都是实现手段,不是独立的用户价值。
每个里程碑必须回答:
- 用户现在可以完成什么此前不能完成的任务?
- 是否形成了真实、可调用、端到端的用户路径?
- 本次新增的基础设施是否被该用户路径实际使用?
如果答案只是“新增了 Schema、Core、测试或治理记录”,但用户仍不能多完成一步,则不得把它表述为产品里程碑。
3. Walking Skeleton / Demo First
本项目默认采用可运行骨架优先,而不是基础设施优先。
正确顺序:
复用已有能力
→ 打通一条真实用户入口
→ 经过最薄的共享连接层
→ 调用真实 Compiler / Evidence / Adapter
→ 产出用户可以检查和使用的 Skill
→ 再按实际风险逐步完善
Demo 不是虚假界面或一次性伪实现。它必须是一条使用真实组件的最小端到端路径,但只覆盖一个最典型场景。
对 Loop Craft,第一条合格 Demo 至少要求:
- 用户可以直接调用
loop-craft,而不是先手写 Accepted Definition JSON; - Agent 能通过已有 From-scratch 访谈形成一个 bounded Loop;
- 用户能审阅并确认关键行为边界;
- 已确认内容进入真实 Compiler;
- 最终得到干净 Skill 与独立 Evidence Package;
- 产物可以被实际检查和调用。
只有 Core、Schema 或单 Loop fixture 的构建链不算 Demo,只算内部技术切片。
以下情形必须停止继续向局部深挖:
- 连续两个开发任务没有增加新的用户可执行能力;
- 测试数量增长,但端到端用户路径没有增长;
- 正在加固的组件尚未被任何可用入口调用;
- 后续架构很可能改变当前抽象,但当前仍在大量投入边界测试;
- 为了让内部模块“完整”而推迟首个可用 Skill。
发生这些情形时,立即回到最短用户路径,不继续追加局部完整性工作。
4. Reuse Before Build
本项目已有高质量实现和参考,默认先迁移、包装或本地化,不从零重建。
资源所有权:
- From-scratch / Craft:优先复用
C:/Users/Administrator/Documents/loopy-skill-handoff/loopy/SKILL.md的访谈、Craft 和预检逻辑。 - Skill-to-Loop Upgrade:优先复用
loopy/references/upgrade-skill.md的 Loopability Gate、四类 verdict、审批和交付流程。 - Discover / Loopability:优先复用
loopy/references/discover.md与现有 SKILL 主体。 - Conversation Distiller:优先本地化
C:/Users/Administrator/Downloads/workflow_skill_creator/SKILL.md的工作流恢复、渐进澄清和严格/灵活步骤分类。 - 既有 Skill 审查与最小优化:优先使用 Skill Polisher。
- 新 Skill 生产化:优先使用 Skill Creator Pro;官方 Skill Creator 只作为 Codex 格式兼容底线。
- Skill 写作:使用 Matt Pocock Inspired Skill Writing Guidelines 作为横切约束。
新增公共模块前必须证明:
- 至少两个真实入口需要同一行为;
- 现有实现不能通过小型包装复用;
- 该抽象会立即被当前端到端路径使用。
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 · 185 lines · 2,622 tokens per session scan A 98bcbc023745
loop-craft AGENTS.md is an instructions file published in the GitHub repository Conradgui/loop-craft (5 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 2,622 tokens to every session, about $0.0131 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
flyto-core AGENTS.md
Instructions for flytohub/flyto-core, covering flyto-core agent rules, flyto2 project memory contract, flyto2 frontend quality gate and repo notes.
flyto-core CLAUDE.md
Instructions for flytohub/flyto-core, covering claude notes, cross-agent handoff and shared code intelligence.
r3 AGENTS.md
Instructions for hyperlogue/r3, covering r3 — review. revise. resolve, architecture, module layout, domain model and review kinds & sources.
briefloop AGENTS.md
Instructions for Stahl-G/briefloop, covering agents.md, purpose, instruction scope, environment separation and context mode.
briefloop CLAUDE.md
Instructions for Stahl-G/briefloop, covering claude.md, claude code role, first response in claude code, standard claude code path and repository development setup.
evidra CLAUDE.md
Instructions for vitas/evidra, covering claude.md, build & test commands, git, architecture and binaries.