Security audit
Security checks for vulnerabilities and agentic risk
This skill is a simple reasoning-mode helper with no code, network, file, credential, or persistence behavior, but users should know it can change reasoning mode automatically on complex prompts.
Install this only if you want an always-available helper that may turn reasoning mode on or off for complex prompts without asking each time. Review the icon convention and explicit controls so you can disable reasoning when speed or token use matters.
The skill is declared to trigger on every user message as a preprocessing step, which makes its scope effectively global rather than limited to clearly defined contexts. Broad always-on behavior can override user intent, interfere with normal response policies, and create unpredictable control flow across unrelated conversations even without external code execution.
The instruction to run a mental complexity assessment on every request is ambiguous and overly broad, causing the skill to insert itself into all interactions without a narrow trigger. That increases the chance of unintended behavior, including unnecessary state changes and conflicts with higher-priority instructions or user preferences.
The skill explicitly directs silent activation of reasoning mode and appending hidden state indicators without asking the user, which is a transparency and consent problem. Silent internal mode switching can bypass user expectations, consume additional resources, and make responses appear behaviorally inconsistent without disclosing that a persistent setting was changed.
No suspicious patterns detected.
