honey-compress

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly instructs preserving exact code/commands and any "load-bearing" directives (including auth/secrets) verbatim when rewriting, which forces the model to copy any secret values from the file into its output.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). This skill explicitly directs the agent to create a backup and rewrite a file that it re-reads every session, performing persistent on-disk modifications that can change agent behavior and thus alter the machine's state even without requesting sudo.