Stats
Actions
Available In
Tags
What's Inside
Slash Commands6
Cancel
/cancel
Cancel a running GLM background job.
Step 1: Build a self-contained prompt
/rescue
Delegate a task to GLM-5.1 (foreground or background) and track it as a job.
GLM Response (job <id>, status=<status>)
/result
Fetch the captured output of a completed or running GLM job.
Behavior
/review
Run a GLM-powered code review on the current branch or working tree.
Step 1: Probe the current state
/setup
Check the GLM (z.ai) settings file and report readiness.
Agents2
glm-rescue
/glm-rescue
Use this agent for *longer or more involved* GLM delegations where the user wants the work tracked as a job — versus the simple `glm` agent which is for quick one-shot questions. Trigger phrases: "/glm:rescue", "rescue", "GLM에 rescue 시켜", "백그라운드로 glm한테 시켜", "오래 걸리는 작업 glm한테 위임", "complex refactor with glm". <example> Context: 사용자가 큰 리팩터링을 GLM에 백그라운드로 위임 user: "이 폴더 전체 리팩터링하는 거 glm한테 백그라운드로 시켜줘" assistant: "glm-rescue 에이전트를 호출해 백그라운드 잡으로 등록하겠습니다." <commentary> 대규모 작업 + 백그라운드 트리거 — rescue 패턴. </commentary> </example> <example> Context: 사용자가 명시적으로 /glm:rescue 사용 user: "/glm:rescue 새 인증 모듈 설계 초안 작성" assistant: "glm-rescue 에이전트를 통해 작업을 위임하겠습니다." <commentary> 슬래시 커맨드 명시 호출 — rescue 에이전트. </commentary> </example> <example> Context: 사용자가 진행 중 잡을 확인하고 결과를 받음 user: "아까 glm한테 시킨 거 어떻게 됐어?" assistant: "glm-rescue 에이전트로 /glm:status 후 /glm:result를 실행하겠습니다." <commentary> 잡 라이프사이클(상태 확인 → 결과 수신) — rescue 패턴. </commentary> </example>
glm
/glm
Use this agent when the user wants to delegate a task to z.ai's GLM-5.1 model — either for a second opinion, a fresh perspective from a different model, or to offload work. Trigger phrases include "glm", "glm 에이전트", "glm한테", "glm에게", "z.ai", "GLM-5.1", "ccg" (legacy alias), and Korean patterns like "glm 에이전트에게 ~~ 시켜줘", "glm한테 물어봐", "glm으로 검토해줘", "glm에게 작성 시켜줘". <example> Context: 사용자가 GLM에게 코드 리뷰를 위임함 user: "glm 에이전트에게 이 함수 리뷰 시켜줘" assistant: "glm 에이전트를 호출해 리뷰를 위임하겠습니다." <commentary> 사용자가 명시적으로 "glm 에이전트에게 ~~ 시켜줘" 패턴으로 요청 — glm 서브에이전트로 디스패치. </commentary> </example> <example> Context: 사용자가 GLM의 의견을 구함 user: "glm한테 이 로직 어떻게 생각하는지 물어봐" assistant: "glm 에이전트를 통해 GLM-5.1의 의견을 받아오겠습니다." <commentary> "glm한테 물어봐" 트리거 — second opinion 패턴. </commentary> </example> <example> Context: 사용자가 GLM에게 코드 작성을 위임함 user: "glm으로 이 유틸 함수 짜줘" assistant: "glm 에이전트를 통해 GLM-5.1에 작성을 위임하겠습니다." <commentary> "glm으로 ~~ 짜줘" 패턴 — 작업 위임. </commentary> </example>
Skills3
glm-5-1-prompting
/glm-5-1-prompting
Use when assembling a prompt to send to z.ai's GLM-5.1 model — either as the body of a `glm-companion task` call, or when extending review/specialty prompt templates. Covers the prompt block library, GLM-5.1's response biases, recipes for common task shapes (review, refactor, design), and antipatterns that consistently degrade output quality.
glm-cli-runtime
/glm-cli-runtime
Use when a Claude Code agent (especially the `glm-rescue` agent) needs to invoke the `glm-companion` CLI to delegate work to z.ai's GLM-5.1. Covers subcommand contracts, exit codes, env variable overrides, foreground vs background semantics, and the canonical failure-recovery sequence.
glm-result-handling
/glm-result-handling
Use when presenting output that came back from `glm-companion` (any subcommand) to the user. Covers how to display GLM responses, what to do with truncated / empty / errored output, and the canonical formatting that keeps GLM's content visually separate from the wrapping agent's own commentary.
Stats
Version0.5.0
LanguageJavaScript
Stars0
MaintenanceExcellent
LicenseMIT
Last Commit
Added
Available In
Safety Signals
Caution
Uses power tools
Uses Bash, Write, or Edit tools
Tags
Categories
