Security audit
Security checks for vulnerabilities and agentic risk
This skill coherently monitors AI token usage from local session logs and stores derived usage summaries, with no evidence of exfiltration or hidden execution.
Install only if you are comfortable letting the skill read local QClaw session logs to compute token totals. It does not appear to transmit data externally, but session logs may contain sensitive conversation history, and the generated usage summaries persist under workspace memory until deleted.
The skill describes reading session logs and writing per-day statistics, but it does not declare any explicit tool scope such as permissions or allowed-tools. That creates an authorization ambiguity where the runtime or user may not clearly understand that local file access is required, increasing the risk of unintended access to sensitive local data if the skill is triggered.
The skill explicitly states that it reads session logs and stores derived statistics, but it does not disclose the privacy implications of analyzing potentially sensitive conversation history or persisting summaries to disk. Users may unknowingly expose prompts, outputs, usage patterns, or other metadata contained in local session logs, making accidental data disclosure more likely.
The trigger phrases include very generic terms like "Token", "费用", and "统计", which can easily appear in normal conversation unrelated to this skill. Overbroad triggers increase the chance of accidental activation, which is especially risky here because the skill is designed to inspect session logs and produce stored statistics from local data.
The module title and user-facing description are written in Chinese, and the script's interface continues in Chinese-only prompts and output. This creates a locale policy issue because the skill does not provide user opt-in or an alternative language despite imposing a specific language on all users.
The natural-language instructions and trigger examples are presented only in Chinese, which can imply a locale-specific requirement or forced language behavior. There is no explicit opt-in, alternative language support, or justification that this skill is intentionally limited to a Chinese-speaking context.
The manifest describes monitoring, reading logs, anomaly detection, and optimization advice, which implies analysis of existing data. The code goes beyond read-only monitoring by creating directories and writing per-day JSON records under ~/.qclaw/workspace/memory, adding a persistence side effect not stated in the manifest.
No suspicious patterns detected.
