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/bifrost-proxy/bifrost/agents-mdgit clone --depth 1 https://github.com/bifrost-proxy/bifrostWhat 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 | $0.16937 | $0.16937 |
| Opus 5 | $0.08469 | $0.08469 |
| Sonnet 5 | $0.03387 | $0.03387 |
| Haiku 4.5 | $0.01694 | $0.01694 |
Grade A, and why
bifrost AGENTS.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 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- **CI/Pages 看护**：推送部署仓库后必须跟进 `bifrost-proxy/bifrost-proxy.github.io` 的 `Pages` 与 `pages-build-deployment` runs 到 success，并用 `curl -L https://bifrost-proxy.github.io/` 验证线上根站已经是源站同步后的首页，不是旧跳转页或手写分叉页。 How it starts
The opening of the file, as written. The whole thing — 698 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Bifrost 项目开发规则
执行心法
任何人都会犯错，但我们可以通过反复检查、持续改进，终将达到巅峰水平。 Agent 必须把 review、修复和复测视为高质量交付的常态，而不是额外负担；任何没有证据支撑的“已完成”“应该可以”“看起来没问题”都不能作为完成依据。
执行模式与完成定义
任务模式判定
- 开发模式（默认）：用户要求实现、修复、优化、改造、补齐、提交、推送，或要求 Agent 自主完成时，必须执行本文的通用交付闭环，并按“变更类型与验证路由”只选择适用的专项流程；除非用户明确说“不要提交 / 不要推送 / 不用建 MR / 不用盯 CI”，否则开发完成后必须默认提交、推送、创建或更新 MR/PR，并跟进适用的远端 CI 到全绿或给出明确阻塞归因。
- 检查模式（只读）：用户明确说“仅检查 / 不要修改代码 / 只 review / 先分析”时，只能读取、运行安全检查并给出 findings；未经用户明确同意不得编辑文件、提交或推送。
- CI 闭环模式：用户要求分析 CI、修到 CI 绿、推送后盯 CI、进入 fix-push-watch，或任一开发任务完成后已 push 需要默认远端 CI 看护时，必须加载
.agents/skills/github-actions-pat/，并按 GitHub Actions CI 分析与 fix-push-watch 闭环执行。 - 文档/流程变更模式：仅修改文档、流程或测试说明时，只执行格式、链接、结构和内容一致性等针对性检查；默认不触发 Rust 编译/测试、覆盖率、E2E 或
human_tests/。
变更类型与验证路由（先判定，后执行）
验证范围必须由实际影响面决定，不能因为“修改了仓库文件”就运行所有测试。任务开始时先根据 diff、依赖关系和运行时影响选择下表中的一类或多类；混合变更取各类适用验证的并集，影响面无法确认时才升级验证范围。
| 变更类型 | 典型范围 | 默认必做验证 | 默认不适用 |
|---|---|---|---|
| Rust 后端/代理/协议生产代码 | crates/*/src/**/*.rs、服务端行为、共享协议 |
受影响 crate 单元测试、Rust fmt/clippy、make coverage-changed、按风险选择相关 E2E；大范围或跨 crate 变更再执行 workspace/local-ci |
无关前端测试 |
| 前端行为/状态/API 集成 | WebUI/desktop 的 TS/TSX/JS 逻辑、路由、状态管理、接口调用 | 前端格式化、类型检查、受影响单元/组件测试；用户链路变化时做相关 UI/E2E 或真实场景验证 | Rust coverage、Cargo workspace、Rust E2E（共享协议或后端同时变化除外） |
| 纯 UI 视觉 | 样式、布局、图标、主题、纯展示文案 | 前端格式化/类型检查、受影响页面的亮暗主题和视觉检查；有现成快照时更新并验证 | Rust coverage/build、仓库级 E2E；无需为纯视觉改动强制新增 human_tests/ 文档 |
| Skill/脚本/工具链 | .agents/skills/、skill 自带脚本、独立工具脚本 |
如提供则运行 quick_validate.py，并执行自带单元/脚本测试、shell/语法检查；涉及外部真实流程时只验证该流程 |
Rust coverage/build、仓库级 E2E、human_tests/（除非修改 Bifrost 产品运行时行为） |
| 文档/流程/配置说明 | Markdown、设计文档、AGENTS.md、说明性配置 |
Markdown/格式、链接、结构、一致性检查和两轮 review | Rust coverage/build、前后端 E2E、human_tests/ |
补充原则：
- 文件后缀不是唯一判断依据；脚本若改变产品启动、发布或协议行为，应按受影响的运行时表面追加验证。
human_tests/只用于自动测试不足以证明的用户可感知运行时行为、真实账号/设备/浏览器链路或高风险回归，不是所有开发任务的默认门禁。rust-project-validate、Cargo workspace 测试和 Rust 覆盖率只在 Rust/后端影响范围内启用；前端、Skill、文档或流程变更不得为了“流程完整”空跑这些检查。- 最终验证矩阵只列适用项；不适用项可按变更类型统一说明，无需逐项制造占位任务。
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 · 698 lines · 16,937 tokens per session scan A 2dd935757818
bifrost AGENTS.md is an instructions file published in the GitHub repository bifrost-proxy/bifrost (124 stars, last pushed 3d ago), licensed MIT. It adds 16,937 tokens to every session, about $0.0847 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 2026-08-30.
Other instructions, from other repositories
prodex AGENTS.md
Instructions for christiandoxa/prodex, covering agents.md, scope and source of truth, project summary and ownership, before changing code and mojo-first development.
nanobrowser CLAUDE.md
Instructions for nanobrowser/nanobrowser, covering claude.md, project overview, development commands, workspace tips and architecture.
claude-tap AGENTS.md
Instructions for liaohch3/claude-tap, covering maintainer automation notes, agents 索引, documentation boundary, review guidelines and pre-commit hook.
hoop CLAUDE.md
Claude Code instructions for hoophq/hoop, covering claude.md, project overview, toolchain & prerequisites, architecture and module breakdown.
octocode AGENTS.md
AGENTS.md instructions for bgauryy/octocode, covering agents.md — octocode monorepo, dogfood, architecture, packages and tools.
thinkrail AGENTS.md
Instructions for JetBrains/thinkrail, covering thinkrail, module structure & boundaries (top-priority requirement), engine: pi only, in-process, architecture (three rings) and repo layout.