agentleFS
buidangminh23buidangminh23/codex-mcp-bridge/skills/codex-bridge/SKILL.md

codex-bridge

Skill32 starsChanged 17 days agoMITView on GitHub

Read saved projects and conversations or delegate authorized work to the user's paired local Codex Desktop through the Codex Bridge plugin.

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 1 is distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this workflow when a user wants ChatGPT to inspect Codex Desktop conversations or send work to their paired computer.

    line 8
  2. Very specific:

    Call codex_bridge_status. If unpaired or offline, direct the user to the local connector's pairing URL or ask them to start their installed …

    line 10
  3. Very specific:

    Call list_codex_projects or list_codex_threads to verify the exact saved project and destination.

    line 11
  4. Very specific:

    Use read_codex_request with the returned queue ID.

    line 14
  5. Very specific:

    After a confirmed send, use read_codex_thread to inspect progress or replies.

    line 15

“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Codex Bridge

Use this workflow when a user wants ChatGPT to inspect Codex Desktop conversations or send work to their paired computer.

  1. Call codex_bridge_status. If unpaired or offline, direct the user to the local connector's pairing URL or ask them to start their installed connector. Never claim that an offline computer received work.
  2. Call list_codex_projects or list_codex_threads to verify the exact saved project and destination. These calls return queue operation IDs. Use read_codex_request to retrieve their results before choosing a destination.
  3. Create a new conversation with start_codex_thread only when the user explicitly or through standing instructions authorizes a new task. Supply the exact saved project path, a clear title, a complete English brief and a fresh UUID requestId. Structure briefs as Goal, Context, Task, Scope, Constraints, Done when and Reply format when useful. Preserve literal user text and file names.
  4. Continue unfinished work in its verified original thread with send_to_codex_thread. Read that thread before sending if its identity or current work is uncertain. Supply a fresh UUID per distinct follow-up, retaining the same UUID only for a retry of that exact follow-up.
  5. Use read_codex_request with the returned queue ID. A queued or claimed operation is not a completed Desktop action. The result of a completed operation can still indicate a blocked or uncertain Desktop action. Read the actual result before reporting success.
  6. After a confirmed send, use read_codex_thread to inspect progress or replies. Desktop acceptance is separate from completion of the user's task.

Never automatically resend work after an uncertain, disconnected or expired claimed operation. Preserve its queue ID, request UUID and any confirmed thread ID. Inspect the original task. An expired_not_dispatched request was not taken by the connector. A disconnected computer does not imply that previously dispatched Desktop work stopped.

Each ChatGPT account controls only its paired computer. Local allowed roots, saved project identity and original signed-in account bindings are checked by the connector. Do not request a broader root, different account, arbitrary native operation or CLI fallback to bypass a rejection.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.