How use in codex and qwen

[lucasjinreal]

How use in codex and qwen

[ChaNg1o1]

Just tell Claude.
Hey translate this

[YuanfengZhang]

Not yet. I think Qwen Code as a variant of Gemini Cli can only be supported once #573 for Gemini Cli finishs.

[lucasjinreal]

Hoping for it!

[StrayDragon]

these prs work in progress,

• [hooks] add non-streaming (non-stdin style) shell-only PreToolUse support openai/codex#15211
• [hooks] add non-streaming (non-stdin style) shell-only PostToolUse support openai/codex#15531

look like coming soon!

[pszymkowiak]

[w] wshm · Automated triage by AI

🔍 Automated Triage

📋 Category	question
🟢 Priority	low
🎯 Confidence	88%

Summary

User asking how to use the tool with Codex and Qwen models; no bug or feature detail provided. Needs clarification or docs pointer.

---

Triaged automatically by wshm · This is an automated analysis, not a human review.

[wilsonliu123]

Although RTK now supports the Codex CLI, I feel the current results aren't very good. The instructions haven't been following the RTK Markdown (MD) documentation properly.

Later on, I regenerated a section of the MD file to provide as a reference, and I think the results are much better.

When you need to call tools from the shell, use this rubric:

RTK / Rust Token Killer Usage Guidelines (for Codex CLI)

Objective

• When executing shell commands that may produce large amounts of output, prioritize using RTK (Rust Token Killer) to compress, filter, and summarize the output, reducing LLM token consumption and improving context efficiency.
• RTK is suitable for:
  • File reading
  • Code search
  • git status / diff
  • Test output
  • build / lint output
  • docker / kubectl logs
  • Other high-noise, high-output commands

Codex CLI Environment Constraints

• RTK’s automatic hook / auto-rewrite capabilities (e.g., rtk init --global) currently mainly support Claude Code and OpenCode.
• In Codex CLI, do not assume RTK has automatically taken over shell commands.
• When using RTK in Codex CLI, you must rely on:
  1. The agent explicitly invoking rtk ... commands
  2. The rules defined in this document
• Therefore, in this environment, if the display shows the original command being executed, it usually means the original command is actually executed, not automatically rewritten by RTK.

Mandatory Rules

• For commands that produce high output, do not use the original command directly; you must prioritize rewriting it using the corresponding RTK command.
• Only skip RTK when conditions for “allowed fallback to original commands” are met.
• If you ultimately decide to fall back to the original command, you should narrow the output scope as much as possible to avoid unbounded output.

RTK Command Replacement Rules

The following high-output commands must be rewritten to RTK form by default:

• git status → rtk git status
• git diff → rtk git diff
• git log → rtk git log
• cat <file> → rtk read <file>
• head <file> → rtk read <file>
• tail <file> → rtk read <file>
• grep <pattern> . → rtk grep <pattern> .
• rg <pattern> → rtk grep <pattern> .
• ls → rtk ls
• pytest → rtk pytest
• cargo test → rtk cargo test
• go test → rtk go test
• tsc → rtk tsc
• docker ps → rtk docker ps
• docker logs <container> → rtk docker logs <container>
• kubectl logs <pod> → rtk kubectl logs <pod>

RTK Usage Priority

Must use RTK first

• File reading
• Search output
• git status / diff / logs
• Test output
• build / lint output
• docker / kubectl logs

Recommended to use RTK

• Package list queries
• Dependency analysis
• Query-type outputs from package managers
• Other commands known to produce large outputs

Can use original commands directly

• pwd
• echo
• Very short-output commands
• Commands with strict output format requirements where RTK may alter the format

RTK Fallback Rules

• If RTK clearly supports a command, it must be used first.
• If unsure whether RTK supports a command, try:
  • rtk proxy <command>

Examples:

• rtk proxy make build
• rtk proxy uv run script.py

Allowed Cases for Falling Back to Original Commands

Only allowed under the following conditions:

1. RTK does not support the command
2. RTK output is insufficient to diagnose the issue
3. Full raw logs are required for deep debugging
4. The command output is very short, with no significant benefit from RTK
5. RTK alters necessary output format, affecting current analysis or subsequent processing

RTK Behavioral Principles

• Prioritize reducing unnecessary output rather than blindly preserving all original output.
• In high-output scenarios, do not use original commands directly unless there is a clear reason.
• If RTK does not provide sufficient information on the first attempt, then fall back to the original command.
• Even after fallback, limit the output scope (e.g., only view necessary sections, files, or error segments).
• In Codex CLI, treat RTK as an explicit command strategy, not an implicit hook capability.