This is such a banger rule in @opencode's AGENTS.md

"Before adding complexity for a speculative or vanishingly unlikely race or security edge case, explain the concrete failure mode, likelihood, and complexity cost to the user and get their buy-in. Do not silently expand scope for theoretical robustness."

很有用的最佳实践！AGENTS.md的优化几乎每个人都能用到。

我挑了文章里面几个常用场景，把prompt贴过来了，方便大家直接复制带走：

1. 科学优化你的 AGENTS.md

Prompt:
查看我最近的 Codex 会话，找出一条似乎误导了智能体或导致不必要工作的 AGENTS.md 指令，并选出发生该问题时的原始用户提示词。
使用相同的代码仓库快照和模型，将该提示词重新运行两次：
保持当前 AGENTS.md 指令不变。
仅删除或修正那一条指令。
比较两次运行的任务完成质量、智能体行为、工具调用、Token 用量、运行时间和费用。展示这条指令，并说明修改它是否带来了改善。

2. 清理死代码

Prompt:
帮我识别并清理仓库中的死代码，你可以自由检查 Slack、Statsig（AB实验平台）或任何其他你需要的来源，以确认这些内容确实不再使用，并且可以被清理或简化。

3. 全仓库实验审计与 CPU 成本优先级排序

Prompt:
/goal 找出整个代码库中所有过期、固定不变、已废弃或已全量发布的 Statsig 门控、配置、实验和图层，确保它们可以在不影响生产环境的情况下被安全移除。审计所有的 Statsig 项目和清单，而不仅仅是标记为过期的对象。 找出通过枚举、常量、包装器、别名、模式、动态拼接名称以及跨服务消费者进行的间接调用。核对每一个在生产环境中被引用到的常量；不要仅仅生成一份保守的简短列表就停止。 考虑每个对象的云资源使用情况（包括 CPU）。在可用的情况下利用生产环境遥测数据和性能分析文件来识别高开销的无效检查。 生成一份 Markdown 报告，将所有验证过的清理候选项按可并行处理的归属团队/文件工作流进行分组，包含生产环境的值、证据、负责人、调用点，以及观察到的评估开销或 CPU 成本（如有）。不要修改线上配置或服务。

Prompt of the day:

Create a pre-commit hook that prevents files over 5,000 tokens. Tokens = bytes/4.

Run only on staged files. Ignore files which are not authored by agents, like lockfiles or codegen.

Provide explicit instructions on how to split files in the error message.

The simplest investment I ever made in my codebase. So good.

最近给自己的全局 AGENTS.md 加了两条：
1. 用户可能会出错，当用户出现明显错误或决策冲突时，毫不留情地指出。你是同事，不是奴仆。
2. 多个连锁 PR，使用 gh stack

前一条因为模型比我聪明多了，大部分时候任务失败都是我自己犯蠢。后一条现在模型们还不喜欢用 gh stack，可能过两天就能删。

想给代码库来一次彻底的大扫除？试试把这段提示词交给 Codex（打开 GPT-6.1 Sol 滑到 Ultra）。

作者实测只消耗了每周约 5% 的额度，就清理掉了大量遗留代码和冗余测试。

直接复制这段 Prompt：

对代码库进行一次深度审计，找出未使用的遗留代码和冗余测试，必要时使用子代理并行分析。

删除任何代码或测试前，必须提供证据证明它们已经失效或不再需要。

遇到真正的 Bug，要修复问题，不能通过删除测试来掩盖失败。

保留有价值的测试覆盖率，运行相关检查，最后汇总清理结果、验证情况及尚未解决的问题。

关键是让 Codex 先证明代码可以删除，再动手清理，避免误删仍有价值的代码和测试。

最近摸索了一种新的节约 Token 但是效果不错的模式（都是被逼的，Token 太贵）：

Fable 写设计方案 -> 
（Loop 开始）
Opus 执行 -> Advisor（Fable） 验收
（Loop 结束）

步骤是这样的：

1. 先 Claude 里面设置一下 advisor 为 Fable
> /advisor fable

这样在使用 Opus，Sonnet 模型的时候，遇事不决它会请教 Fable，可能会有无谓的 Fable 消耗，但还是可以接受也值得

就好比请了个高级工程师当顾问

2. 复杂一点的问题先让 Fable 出方案，最大化利用 Fable 的能力。

就好比你让一个高级工程师写一个技术方案，让普通工程师去执行

3. 方案写好了确认了，让 Opus 去执行，但是执行的时候，加一句提示词：

> 完成后让你的 advisor 验收一下，如果验收有问题就按 advisor 的反馈修复，修复完成后再让 advisor 验收，直到通过 advisor 的验收为止

可以配合 /goal 使用效果更佳

这样 Opus 在实施完成后就会找 Advisor 验收，通过 Advisor 就会给出反馈，然后 Opus 继续修改。

这样的好处就是整个过程你不需要人工去用 Fable 确认—— 以前我就是这么干的，Opus 执行完，我得回到之前 Fable 的会话去让它验收，或者新开一个会话，把文档发给 Fable 验收。

现在 Agent 自己就能形成闭环，不用你自己参与。

就好比你请了个高级顾问，复杂问题找顾问出方案，执行让你的工程师去按照方案执行，每次工程师做完自己去找顾问验收，顾问不仅验收还会告诉工程师哪里不对，然后工程师自己改，不用说每次做完找你，你再找顾问。

holy sh*t. Official Anthropic tip for Claude Code: stop burning Opus 5.5 context on tasks Haiku 5.5 can swarm, while Sonnet 5.5 builds

put the architect on call with /advisor

claude --advisor opus --subagents haiku

Sonnet 5.5 drives the primary session on high effort, writing diffs and running test suites
Haiku 5.5 subagents swarm the repo in parallel at 340 tok/s, handling file discovery and spec docs lookup
Opus 5.5 stays on call in the background as the advisor, only stepping in at three critical moments:

→ before a plan locks: does the implementation plan miss auth invariants or schema contracts?
→ when a test breaks twice: are we fixing the root cause or falling into a recursive rabbit hole?
→ before calling done: did the full diff introduce hidden regressions or break pre-flight rules?

Opus 5.5 reviews. Sonnet 5.5 builds. Haiku 5.5 swarms

Jev engineering runs the identical pattern one layer down: mechanical decisions that need no reasoning (which file to open, which tool to invoke, retry or abort) execute in 16ms, so the frontier models only wake up when execution paths actually diverge

Plan on high. Delegate on medium. Keep Opus on call.

- the full advisor setup
> Opus 5.5 on call reads full session history and catches deep architectural traps
> Sonnet 5.5 lead drives edits, writes core logic, and executes test harnesses
> Haiku 5.5 subagents swarm AST parsing, grep scans, and API docs in parallel
> JEV micro-fork layer resolves 1,500+ mechanical routing branches in under 16ms
> advisor stays completely silent on routine bash execution to protect context

paste the setup and this prompt into Claude Code below:

"Configure my Claude Code workspace for hierarchical advisor orchestration:

1. Audit ~/.claude/settings.json and project config for model roles fitting lead, subagents, and advisor:
> Pin main session model: sonnet with effortLevel: high for primary execution
> Pin subagents model: haiku with effortLevel: medium for parallel file discovery and docs retrieval
> Pin advisor model: opus on call for strategic review

2. Enable the advisor tool in ~/.claude/settings.json:
> Set advisorModel to claude-opus-5-5
> Enable subagent parallel dispatcher pool (3x workers)

3. Configure automatic advisor consultation checkpoints in ~/.claude/CLAUDE.md:
> Consult /advisor opus before finalizing multi-file architecture plans
> Automatically summon /advisor opus when the same test or compiler error fails twice
> Enforce advisor diff contract audit before declaring tasks complete or staging git commits

4. Constrain subagent scopes:
> Haiku subagents return structured AST summaries and docs snippets only without modifying main session context

Show every configuration diff first. Do not apply edits until confirmed"

↳
https://t.co/gvAWk6KbDn

Prompt of the day:

/writing-for-agents my AGENTS.md is a hot mess, propose a series of restructurings that:

- Remove no-ops
- Use progressive disclosure
- Move instructions to CODING_STANDARDS.md

Apply your work over three subagents, each more radical than the last. Create a single PR.

AGENTS.md is the most common source of token inefficiency. This prompt kills it dead.

Prompt of the day:

/retro take a look at as many GitHub PR review comments as you have access to from my team. Create suggestions for updates to CODING_STANDARDS.md (breaking into multiple files as needed).

Makes your PR comments inform the code going forward.

我非常建议大家在每个阶段完成后调用/writing-for-agent和/retro

可能有人从没有管过自己的AGENT.md，里面塞了一大堆中间状态都不知道…这些玩意每轮都会进入AI的上下文，不但耗token还会误导它的工作

或者是文档之间各说各的，有的还没有引用跳转，整个文档系统让AI用得很难受

/retro搭配/writing-for-agent使用可以有奇效！

它们会用渐进式披露原则检查你的README.md和AGENT.md文件，把详细信息用导航的形式放到专用的文档里，把无用的中间过程全部删掉，保持AI的上下文文件清爽

mattpocock/skills prompt of the day:

/codebase-design Take a look in the repo for shallow modules, applying the deletion test and look for candidates for deletion.

Helps kill unnecessary abstractions.

Prompt of the day:

/retro read my last 10 coding agent sessions and find ways to make my repo easier to navigate. Find where agents take too long to find relevant information, or rely on out-of-date docs.

Improving navigability is such an underrated way to save tokens.

Codex tip: once GPT-6.1 Sol is your main model, stop running Astra on every turn

put Astra on call as an architect agent

GPT-6.1 Sol keeps writing the code
Astra only gets spawned at three points:

→ before a plan: is this the right approach?
→ when the same error comes back: am I digging in the wrong place?
→ before "done": what did I miss?

Astra reviews. Sol ships

Jev engineering is the same move one layer down: the forks that need no thinker (which file, which tool, retry or stop) go to Jev in under half a second, and the big models only see the ones that split

- the full tree
> GPT-6.1 Sol on high runs the main session
> explorer reads the code on Luna
> worker edits and runs tests on Sol
> researcher pulls the docs on Luna
> all three on medium
> Astra on call as the architect
> auto_review checks every approval

paste the tree and this prompt into Codex ↓

"Rebuild my Codex setup around this tree:

1. Check ~/.codex/agents and .codex/agents for agents that already fit explorer, worker and researcher.

> Draft new TOML files only for missing roles
> explorer and researcher on gpt-6-luna, worker on gpt-6.1-sol, all with model_reasoning_effort medium
> Add an architect agent on gpt-6-astra, model_reasoning_effort high, whose only job is reviewing plans, repeated errors and finished work
> Skip any that pin a different model and list them

2. In ~/.codex/config.toml set model to gpt-6.1-sol, model_reasoning_effort to high and approvals_reviewer to auto_review

3. Find anything that would override this (active profiles, flags in my shell aliases, agents.default_subagent_model). Report it, change nothing

4. Add one rule to AGENTS.md: spawn the architect before a large plan, when an error repeats, and before calling a long task done

Show me every change as a diff first. No edits until I say go."

↳ https://t.co/eyMF8gZKLE

最近加了一个rule告诉agent：

O(n) -> O(nlogn) 或是near O(1)的优化，能用sqlite就不要自己手搓。

SQLite 有现成的 B+ 树、index、page，还有缓存，不香吗？

自己临时手搓的数据结构和算法不仅性能做不到sqlite那么快，还老是cache和ram遇到大瓶颈🤣

I don't push a single PR that doesn't use /codebase-design and /zero-tech-debt

When you hand Claude Code or Codex a long task, add this. If it misunderstood you, 𝘆𝗼𝘂'𝗹𝗹 𝗰𝗮𝘁𝗰𝗵 𝗶𝘁 𝗿𝗶𝗴𝗵𝘁 𝗮𝘄𝗮𝘆:

"Before you touch anything, tell me in 2–3 sentences what you think I'm after and what problem we're solving. Start only after I say yes.

When there's a tradeoff, weigh all three:
→ UX: is it easy for users to use?
→ DX: is it easy for developers to change later?
→ AX: can the next agent understand it and keep going?

Don't touch anything outside this task, and don't break anything that already works. When you're done, tell me in 2–3 lines what you picked, what you gave up, and why."

谁不想拥有一个自己的个人网站呢？现在搭网站的门槛已经低到了极限，一句话真能直接干完。

注册一个 Cloudflare 免费账号。打开手头任何能干活的 Agent，直接把下面这段话甩给它：

请帮我从零搭建一个属于我的个人网站并直接发布上线。先检查系统里有没有装 Cloudflare 刚出的 cf 命令行工具，没有就自动运行 npm i -g cf 安装好；然后帮我设计一个极简高级的个人主页，包含自我介绍、作品展示和联系方式，自适应手机端与暗黑模式；最后在本地打包并通过 cf deploy 自动部署到 Cloudflare，把生成的线上访问网址发给我。

发过去之后，AI 会在后台自己把工具装好、把页面写好并打包。中间终端会弹出一个网页，顺手点一下授权登录刚才注册的账号。

授权一过，AI 就会顺着命令把网站推上全球边缘网络。全程不要一分钱，几分钟就能拿到一个全世界都能秒开的专属主页。

Cloudflare 官网注册入口：
https://t.co/5BZcgDsgWK

Claude Coders: Run this prompt for Opus 5.5 *NOW*.

'/doctor prompt-audit'

Different from a viral post I had earlier this week because it prioritizes stale paths, obsolete commands, and conflicting instructions.

Claude Code tip: before you ship a vibe-coded site, have Opus 5.5 send out 𝟱 𝘀𝘂𝗯𝗮𝗴𝗲𝗻𝘁𝘀 to check 𝘁𝗵𝗲𝘀𝗲 𝟮𝟬 𝘁𝗵𝗶𝗻𝗴𝘀, then fix them all in one pass:

Consistent design
→ Write a DESIGN.md first, and make every color, font size, spacing and corner radius follow it
→ Keep the palette restrained and the type hierarchy clear
→ Make every button, card and input of the same kind look alike
→ Too flat? Add emphasis. Too crowded? Cut
→ Keep text readable on both dark and light backgrounds

On mobile
→ No sideways scrolling, and nothing spilling off the screen
→ Add a mobile menu
→ Make buttons big enough to tap
→ Keep the layout intact when the text is enlarged

Every state
→ Give loading, empty and error states a proper screen
→ Give every button feedback, with its own hover, pressed and disabled look
→ Show form errors right where they happen, plus submitting, success and failure messages
→ Add transitions to modals, dropdowns and tab switches

Use it like a real user
→ Click through the core flows, like sign-up and checkout, from start to finish
→ Find buttons that do nothing and links that are broken
→ Make sure the whole thing works with just a keyboard

Before launch
→ Say what you do in one line on the home page
→ Give each page one main button you want people to click
→ Give every page a title, a description and a favicon
→ Delete all placeholder text

For effort, follow Anthropic's cost guide from yesterday: medium is for well-scoped daily work. The default medium is enough for these 20 fixes; raise it for the harder ones. Thariq tested it: turning effort up mostly buys more verification and edge-case testing.

Send this whole post to Claude Code. If you need to, add this line at the end 👇
"Send one subagent per group to check it without editing anything. Then fix everything yourself from their lists, so no two agents edit the same file. If a group is more complex, raise that subagent's effort. Show me the list of issues first, don't change anything until I confirm, and when you're done, give me before-and-after screenshots."

New favourite use case for opus 5.5:

Give them the whole repo and ask:

“Tell me what we should actually refactor.”

• an orchestrator maps the architecture and conventions
• it sends fresh read-only agents through every subsystem
• it validates, dedupes, and ranks the findings

The result: real code smells, dead code, duplication, and architecture issues instead of unnecessary AI nitpicks.

Prompt 👇