Security audit
Security checks for vulnerabilities and agentic risk
TokenGuard is a local quota-tracking helper whose file writes are purpose-aligned and do not store prompts or send data elsewhere.
Before installing, understand that this skill is a local quota helper: it may write a state.json file containing model usage counters, and it should be integrated deliberately before API calls rather than assumed to enforce provider limits automatically.
Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.
return self._init_state()
return data
except (json.JSONDecodeError, OSError):
# Corruption detected, reset state
return self._init_state()
return self._init_state()
This code estimates tokens from the provided prompt text and then writes usage state to a local JSON file via _save_state_atomic. Although the state currently stores counts rather than raw prompt content, the script accepts user prompt text as input and performs filesystem writes without any visible warning, confirmation, or user-facing disclosure in code comments/messages about creating and updating state.json.
No suspicious patterns detected.