Security audit
Security checks for vulnerabilities and agentic risk
This skill is a local quantified-progress tracker with disclosed local storage, but users should understand it can automatically keep behavioral and task history.
Install only if you want local, ongoing tracking of interaction-derived progress and task metrics. Review or disable auto_track if you do not want automatic accumulation, and avoid exporting all data unless you are comfortable sharing the full local history.
The skill describes ongoing storage of conversation-derived metrics and history in a local file, but does not present a clear up-front warning or explicit consent flow before collection begins. Because the stored data includes behavioral history, task outcomes, learning records, and achievements over time, this creates a privacy risk through silent persistence and retrospective profiling.
The README presents all usage instructions, trigger phrases, and examples in Chinese, which effectively forces a specific language for interaction. There is no indication that users may opt into another language or that the skill is intentionally limited to a Chinese-speaking region or compliance context.
The documented trigger phrases are very broad everyday language such as asking how today went or requesting a report, which can cause the skill to activate in normal conversations without clear user intent. In a skill that performs auto-tracking and data reporting, this increases the chance of unintended collection, persistence, and disclosure of behavioral or task data.
The README describes local storage and an auto_track option, but it does not clearly and prominently warn users up front that ongoing behavioral/task data may be automatically recorded over time. This weak notice and consent model can lead users to enable or use the skill without understanding the privacy implications or retention of accumulated personal/workflow data.
The trigger conditions are very broad and can activate on ordinary conversation about progress, status, or summaries, causing the skill to run in situations where the user did not clearly request persistent quantification or tracking. In this skill’s context, overbroad activation increases the chance of silent collection, inference, and storage of interaction-derived data without meaningful consent.
The skill intentionally creates a persistent data file under the user home directory and updates it over time, establishing durable session persistence for conversation-derived state. While not inherently malicious, this persistence becomes security-relevant here because it stores profiling-style historical data and expands exposure in the event of local compromise, accidental sharing, or unsafe export behavior.
首次运行时创建数据文件：
# 创建数据目录
mkdir -p ~/.openclaw/data
# 初始化数据文件
cat > ~/.openclaw/data/quantified_self.json << 'EOF'
The workflow explicitly instructs the system to continuously record task, learning, recognition, and goal data during conversation, building a persistent interaction log that can later be surfaced in reports. Even without explicit network exfiltration, accumulating this data increases the blast radius of accidental disclosure, prompt-triggered over-sharing, or later unauthorized access to the local store.
The '导出所有数据' feature can disclose the full accumulated local record, including historical interaction-derived metrics and possibly sensitive behavioral summaries, in a single response. In a conversational system, bulk export is especially risky because a vague or maliciously induced prompt could trigger disclosure of far more data than the user realizes.
The description field is written only in Chinese, and the remainder of the file's explanatory content also uses Chinese exclusively. For a general-purpose data template, this imposes a specific language/locale without any visible opt-in, alternative language option, or documented region-specific justification.
No suspicious patterns detected.