Security audit
Security checks for vulnerabilities and agentic risk
This skill is a local decision-support memory system that clearly centers on storing decision history and preferences, with privacy considerations users should understand before enabling it.
Install only if you are comfortable with the agent keeping local Markdown records of your decisions, preferences, feedback, and inferred risk style under ~/soul-decision. Avoid storing sensitive personal, financial, legal, or business-confidential decisions unless you manage file permissions, backups, retention, and deletion yourself.
The core design of the skill is to persist user decision patterns, risk preferences, and prior inputs in memory files for future reuse. While this is functional for the product goal, it creates a real data retention and exposure risk because these records may contain sensitive personal or strategic information that can later be surfaced or accessed locally.
The skill includes shell commands that create and populate persistent files under ~/soul-decision without an explicit warning that local disk writes will occur. This can surprise users, leave residual personal data on disk, and cause unintended persistence of sensitive decision history or preferences.
The skill describes automatic storage of learned preferences, feedback, and decision records derived from conversation content, but it does not provide a clear privacy notice, retention policy, or consent flow. This creates a risk of collecting and retaining sensitive behavioral data without informed user awareness.
The skill explicitly instructs the agent to infer user risk preferences from conversational phrases and automatically write those traits into persistent memory. Inferred-profile storage is sensitive because it captures behavioral and psychological attributes the user may not realize are being recorded, increasing privacy and misuse risk.
The skill directs automatic capture and persistence of the user's framework preferences based on conversational cues, creating a profile of the user's decision-making style. Even if intended for personalization, this is sensitive longitudinal data that can expose habits, priorities, or vulnerabilities across sessions.
The skill allows retrieval and display of accumulated decision style and risk profile data from prior interactions, which can expose personal or strategic information if accessed in the wrong context or by another local user. Because the underlying data is persisted to local files, the exposure risk is amplified by long-term storage and discoverability.
The README explicitly states the skill is '完全中文化表达，适配中文用户工作流', which indicates a language-specific design choice. Elsewhere the document does not offer multilingual operation, user opt-in, or a justified region-specific compliance reason, so this is a natural-language locale policy concern.
