/zai:glm)
Self-Evolving Skill: This skill improves through use. If the
zaiinvocation drifted, a flag changed, or the passthrough broke, fix this file immediately — don't defer. Only update for real, reproducible issues.
Run the zai CLI to consult GLM-5.2 with the user's arguments, then relay the answer
(note it's GLM's, not Claude's).
zai chat $ARGUMENTS
- If the user passed
--deep/--effort, keep them (extended reasoning). Otherwise it runs fast. - For a big document, they can pass
--file <path>(GLM-5.2 accepts ~1M input tokens). - For web research, use
zai websearch "<query>"/zai read <url>instead. - Full capability surface: the
zaiplugin'sreferences/CAPABILITIES.md.
This is the manual surface (
disable-model-invocation: true) — it fires only when the user types/zai:glm. When Claude should consult GLM on its own initiative, that's theask-glmskill, not this one.
Post-Execution Reflection
After running this skill, reflect before closing the task:
- Did the
zaicall fail or a flag not behave as documented? Fix the command or the argument-hint here immediately, and note it in the zai plugin'sreferences/CAPABILITIES.md. - Did the user reach for
/zai:glmwhenask-glmorzai websearchfit better? Sharpen this skill's description so the right surface triggers next time.
Do NOT defer. The next invocation inherits whatever you leave behind.