glm-review
glm-review
Review a spec, plan, or implementation using Zhipu AI's GLM-5.1 model via the
api.z.ai/api/anthropic Anthropic-compatible endpoint. Runs the already-installed
claude CLI in a child process with three environment variables overridden so the
tool calls land on Z.ai instead of Anthropic.
Designed to run in parallel with codex-review so each review phase gets two
independent model opinions. The consensus matrix (see codex-review §8) extends
naturally to three reviewers.
No fallback. If Z.ai is unreachable or the API key is missing, the review is skipped with a warning and the calling skill continues without GLM input. This is intentional — the user's z.ai quota is open-bar, so failures here are always configuration/network issues, never quota exhaustion worth retrying with a different model.