Security audit
Security checks for vulnerabilities and agentic risk
Token Cockpit locally analyzes OpenClaw usage logs for cost estimates, with no evidence of exfiltration, persistence, or unsafe mutation.
Install this only if you want an agent to read local OpenClaw usage logs and summarize model spend. Review which log path is used, prefer passing --logs explicitly for sensitive environments, and treat dollar amounts as estimates unless you supply current pricing.
Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.
rules.append((k.lower(), (float(v["input"]), float(v["output"]))))
else:
rules.append((k.lower(), (float(v[0]), float(v[1]))))
return rules
return DEFAULT_PRICES
The skill instructs the agent to read local usage logs and auto-detect common filesystem locations and environment-derived paths, but it declares no explicit tool scope or permissions boundary. That creates an avoidable least-privilege failure: an agent may access local files or environment-provided paths more broadly than users expect, increasing the risk of unintended disclosure of local usage data or adjacent files if path handling is loose.
The 'When to use this' section says 'Any time cost or usage comes up' and lists several common conversational phrases such as 'how much am I spending' and 'why is my bill so high.' This is broad enough to overlap with ordinary discussion and does not provide exclusion conditions or context limits, increasing the chance of unintended invocation.