glm5

SUSPICIOUS: the stated purpose is broadly coherent, and the referenced GLM-5 model/API is officially documented, but the skill's real behavior hinges on an unseen local shell script and unspecified API/auth flow. Main risks are unverifiable execution path, outbound model data handling, plaintext reasoning storage, and prompt-injection exposure from reviewer/researcher use with Bash/Write enabled.