error-fixer

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The skill’s workflow is triggered by the user sharing/pasting arbitrary error text (errors, stack traces, console/build failures), which the runtime LLM must ingest directly as free text to interpret and fix.

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 recommends running a privileged command ("sudo chown -R $USER ~/.npm" / "use sudo") which requests elevated privileges and modifies file ownership on the machine, directly altering system state.