vi-humanizer

SUSPICIOUS: the core spec-generation behavior is coherent and mostly local-file oriented, but the skill automatically executes extension hooks from project config and inherits the trust of whatever extensions are installed. The main risk is transitive execution of third-party hook commands, not the spec-writing logic itself.

SUSPICIOUS. The skill’s main behavior matches its stated purpose, and the flagged credential-access strings are documentation false positives. The key risk is transitive execution: repo-controlled extensions.yml can cause automatic hook commands to run, expanding the skill’s footprint beyond simple task implementation. No clear credential theft, third-party proxying, or malicious exfiltration is present, but the command-execution surface is broader than a minimal implementation helper.

SUSPICIOUS: the planning purpose mostly matches the file reads/writes and repo-local setup script, but automatic execution of hook commands from project configuration expands the skill beyond simple planning. There is no clear malware, credential theft, remote exfiltration, or external installer in the provided skill, yet the dynamic hook execution path makes the trust boundary too broad for a low-risk classification.