CtrlK
BlogDocsLog inGet started
Tessl Logo

chatgpt-web-bridge

通过 chatgpt-web 桥(MCP / REST)把本地 agent 接入 ChatGPT 网页版:网页端当大脑,本地只做传输与执行。 触发:「推进/继续网页端会话」「让网页端设计、本地执行」「协作/对等/共同推进」「对等模式/让两边辩一下/协作评审」「共同协作/本地跑任务它维护模型」「用网页端的资料/只读拉取」「把网页版接进 harness」「conversation_id 续接」「网页版比较聪明所以用它」。 不适用:让网页端直接执行本地 shell/文件操作——桥只承载文本,网页端永远没有本地工具。

71

Quality

86%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Low

Low-risk findings worth noting

SKILL.md
Quality
Evals
Security

Quality

Content

77%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is highly actionable and workflow-clear, with concrete tools, endpoints, and validation loops per mode. Its weaknesses are redundancy between the 反模式 and rules sections plus a monolithic structure that does not progressively disclose detail into reference files.

Suggestions

Dedupe 反模式 against 硬规则/循环纪律—keep one as rules and let the other list only the distinct failure modes, or merge them.

Move dated incident notes (2026-09-11/12 实证) into a short '已踩坑/旧坑' or '实证记录' section so they don't pad the live guidance.

Split the C-mode 协议 and 运维 details into reference files (e.g. references/collaboration-protocol.md, references/ops.md) and link from SKILL.md to improve progressive disclosure.

DimensionReasoningScore

Conciseness

Content is operational and assumes Claude's competence, but the 反模式 section restates 硬规则/循环纪律 and dated inline notes (「2026-09-11 定型」「2026-09-12 实证」) pad rather than earn their place; could be tightened. Not a 4 because redundancy is noticeable, not a 2 because no basic-concept padding exists.

3 / 5

Actionability

Fully concrete guidance: named tools (chat_completion, get_conversation, list_conversations), endpoints (REST /v1/chat/completions, 127.0.0.1:9222, /backend-api/conversation/{id}), env vars (W2A_ENABLE_DESTRUCTIVE, W2A_VENV, W2A_PACE_*), commands (start.ps1, chatgpt-web2api ensure), and a copy-ready envelope template.

5 / 5

Workflow Clarity

Each mode (A–D) has numbered steps, explicit 终止条件, and validation checkpoints (核实尾部 before resend, 空结果先核 ID, destructive requires W2A_ENABLE_DESTRUCTIVE=1) with feedback loops for error recovery.

5 / 5

Progressive Disclosure

Good section headers and one clearly signaled external reference (examples/install-other-devices.md), but the ~160-line body is a monolith with protocol/运维/反模式 detail inlined that could live in separate reference files. Not a 4 because no bundle files exist to offload detail and references are sparse.

3 / 5

Total

16

/

20

Passed

Description

95%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is strong: it states the capability, packs in natural trigger phrases with synonyms, and draws a clear applicability boundary. Third-person declarative voice is maintained throughout. Minor room to enumerate concrete actions more exhaustively for a perfect specificity score.

DimensionReasoningScore

Specificity

Names the domain and several concrete actions—「传输与执行」「只读拉取」「网页端当大脑」—across multiple modes, with only minor coverage gaps; not a full 5 because actions are role-level rather than an exhaustive enumerated list.

4 / 5

Completeness

Explicitly answers both what (bridge local agent into ChatGPT web; web as brain, local as hands) and when (dedicated 触发 clause with concrete phrases plus a 不适用 boundary).

5 / 5

Trigger Term Quality

Comprehensive natural trigger phrases with synonyms—「推进/继续网页端会话」「协作/对等/共同推进」「网页版比较聪明所以用它」「conversation_id 续接」—matching phrases users would actually say.

5 / 5

Distinctiveness Conflict Risk

Clear niche (chatgpt-web bridge) with distinct triggers and an explicit 不适用 boundary ruling out local-shell execution, minimizing overlap with other skills.

5 / 5

Total

19

/

20

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository
ooooooooooooooooooop/personal-ai
Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.