careful-delete

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly requires recording and outputting the "exact command" and captured output after execution, which would force the LLM to include any embedded secrets (API keys, tokens, credentials passed as CLI args or headers) verbatim if present.

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

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs the agent to execute destructive system- and data-modifying operations (rm -rf, DROP TABLE, force push, terminate instances, etc.) — i.e. to change/erase machine or service state — even though it wraps them in safety checks, so it enables potentially compromising actions.