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 skills add ooooooooooooooooooop/agent-tools --skill chatgpt-web-bridgegit clone --depth 1 https://github.com/ooooooooooooooooooop/agent-toolsWrote 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/skills/ooooooooooooooooooop/agent-tools/chatgpt-web-bridge)<a href="https://agentmods.dev/skills/ooooooooooooooooooop/agent-tools/chatgpt-web-bridge"><img src="https://agentmods.dev/badge/skills/ooooooooooooooooooop/agent-tools/chatgpt-web-bridge/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/ooooooooooooooooooop/agent-tools/chatgpt-web-bridge"><img src="https://agentmods.dev/badge/skills/ooooooooooooooooooop/agent-tools/chatgpt-web-bridge.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00144 | $0.03183 |
| Opus 5 | $0.00072 | $0.01591 |
| Sonnet 5 | $0.00029 | $0.00637 |
| Haiku 4.5 | $0.00014 | $0.00318 |
Grade A, and why
chatgpt-web-bridge 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 today.
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 — 149 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ChatGPT Web Bridge
快速使用（自然语言即模式）
| 你说 | 模式 | 本地做什么 |
|---|---|---|
| 「推进网页端 XXX 项目/会话」 | A 推进 | 发「继续」式最小提示往前推；纠偏算在内（只给事实不给方案）；判断何时真结束，没结束就不停 |
| 「让网页端 GPT 指挥现在这个项目」 | B 指挥循环 | 它出指令 → 本地执行 → 回帖事实到同一 conv → 直到它说完成 |
| 「和网页端 GPT 协作/对等讨论这个项目」 | C 对等 | 盲答 → 逐条交锋 → 收敛或分歧报告 |
| 「用网页端 XXX 的资料/看它怎么说的」 | D 只读 | 只拉取不发送：会话记录、项目文件、记忆——零写操作 |
会话定位：指明 conversation_id/标题就精确命中；没指明 → 按项目名归类（list_conversations + list_projects 匹配）；仍不确定 → 先问用户再发，不猜。首轮新会话：用户说了项目就带 project_id（名字即可），没说 = 独立会话。
核心原则：谁思考
网页端是大脑，本地是手和邮差。
- 网页端持有完整上下文（会话历史 + Project 指令 + Memory）——它的判断依赖这些，不需要在消息里复述给它。
- 本地侧只做三件事：把文字送进正确的 conversation_id、把回文拿回来、（指挥模式下）忠实执行网页端给出的任务。
- 不要替它写执行步骤。 发「继续」式最小提示，让它自己决定下一步。消息写得越细，网页端思考得越少。
适用边界与安全
触发：推进/继续已有网页端会话、网页端设计本地执行的回合制循环、conversation_id 专线续接。
不适用：
- 让网页端直接执行本地命令或读写文件——桥是文本桥，网页端永远没有本地工具
- 高频自动轮询——这是低频人工节奏
- Chrome 专用 profile 未登录 ChatGPT——先登录再调用，别让请求悬死
安全与验证：所有写操作（新建会话、删除会话）都是真实网页操作；删除类工具需服务端显式 W2A_ENABLE_DESTRUCTIVE=1。本地侧只验证 conversation_id 续接正确、tab 复用正确，不替网页端校验其输出内容的正确性——那是它自己的责任。
循环纪律：本地侧是唯一的发动机
网页端永远不会主动发起下一轮——本地收到回帖不等于回合结束。每轮拿到回文后，先对照本模式的终止条件判断；不满足就继续发下一轮，直到终止条件命中或用户叫停。「它回完了」不是停止理由，「终止条件命中」才是。
四种用法
A. 推进模式（默认，低频）
把已有网页任务往前推一格：
- 最小提示：
继续/按你的结论推进/下一步 - 纠偏只给判断不给方案：
这条线判断错了，事实是 X—— 让它自己想修法 - 每轮必须带 conversation_id；不知道就先
list_conversations找目标 - 网页端产出长报告时，本地只保存/转发原文，不做二次「翻译」
- 终止条件（满足其一才停）：网页端书面宣布本阶段完成且没有它自己的下一步；它提出需要用户/外部世界才能回答的问题（权限、要买的数据、要你拍板的方向）；用户叫停。它写完一轮报告不是终止。
B. 指挥循环（网页端指挥本地）
网页端设计 → 本地执行 → 回帖 → 网页端继续设计：
- 首轮
chat_completion（可带project_id让会话落进项目），存下返回的conversation_id作为专线 - 每轮本地回帖只写事实：结果摘要 + 关键证据 + 阻塞项，不夹带自己的分析
- 网页端输出的指令照做，做完把结果发回同一 conversation_id
- 循环频率低是正常的——这是回合制协作，不是轮询系统
- 继续义务：回帖后必须读它的下一条指令并执行——一轮就停等于循环没建起来
- 终止条件（满足其一才停）：网页端宣布任务完成/无下一轮指令；连续两轮它只复述已有结论不产生新指令；它的指令越权（要本地没有的能力）且澄清后仍无法执行；用户叫停
C. 对等协作模式（手动触发，标签 PEER）
本地模型与网页端同等身份共同推进，没有谁指挥谁。适用前提：token 不缺、本地智力被认可、两边互有知识盲区——网页端看得见它的搜索/项目/Memory，本地看得见文件系统/执行结果/实时状态。任何其他场景回到 A/B。
触发：只在用户明确要求时启用（如「对等模式」「让两边辩一下」「协作评审」）。不自动进入。
What ships with it
6 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- today First seen · 149 lines · 144 tokens per session scan A 6f8d097cf1b6
chatgpt-web-bridge is a skill published in the GitHub repository ooooooooooooooooooop/agent-tools (3 stars, last pushed yesterday), licensed MIT. It adds 144 tokens to every session and 3,183 once invoked, about $0.0007 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 2026-09-12.
Other skills, from other repositories
mix-compression
Reduce mix output noise (5-15% token savings) by installing rtk filters that compress mix test/credo/dialyzer/compile output before it reaches Claude. Use when long mix output floods context.
phx-mix-compression
Reduce mix output noise (5-15% token savings) by installing rtk filters that compress mix test/credo/dialyzer/compile output before it reaches Claude. Use when long mix output floods context.
quick
Implement small Phoenix changes without planning — add validations, update routes, fix components, create migrations. Use for single-file edits under 50 lines.
phx-freeze
Apply an advisory edit scope in this session. Use for read-only or directory-scoped work; no enforcement hook is installed.
042-planning-openspec
Use when creating or updating OpenSpec artifacts from an issue, plan, approved design, ADRs, existing OpenSpec, or a valid combination. The workflow assesses reviewable scope, records source authority and derivation, handles conflicts, and prevents silent synchronization. Triggers include Create OpenSpec from an…
043-planning-github-issues
Use when you need GitHub CLI (gh) installation/authentication guidance and an operator-only GitHub issue inventory workflow. The agent does not ingest GitHub issue, milestone, body, comment, title, label, or summary text; requirements analysis must use repository-owned planning artifacts, with issue numbers only for…