智能自动记账

• Outcome: Review
• Latest audit: Jun 14, 2026 at 5:39 PM UTC
• Version:
  1.0.0

Overview

This bookkeeping skill is not malicious, but it should be reviewed because it can record, reveal, report, and delete personal financial data from broad conversational triggers.

Install only if you are comfortable with an agent managing a local personal finance ledger. Use explicit commands, review parsed entries before saving, and avoid relying on the delete/report flows unless you add confirmation or backup behavior. The package also appears incomplete because it references Python scripts that were not included.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Vague Triggers

Medium

• Confidence: 94% confidence
• Finding: The skill advertises very broad trigger phrases such as '买了', '收到', '支出', and '收入' that commonly appear in ordinary conversation. In an agent environment, this can cause accidental invocation and unintended financial data capture or actions when the user did not intend to use the bookkeeping skill.

Vague Triggers

Medium

• Confidence: 95% confidence
• Finding: The instruction to parse '任何类似记账的话' is overly permissive and leaves invocation boundaries undefined. This increases the chance that unrelated user messages are interpreted as bookkeeping commands, leading to mistaken records and possible follow-on actions without clear user intent.

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: Query examples like '看看' and generic mentions of '账单' or spending can overlap with normal conversational language. This may expose personal financial records in contexts where the user was speaking generally rather than requesting a ledger lookup.

Vague Triggers

Medium

• Confidence: 90% confidence
• Finding: Report-generation triggers like '报告', '月报', and '可视化' are vague and likely to collide with unrelated requests. Because report generation creates and surfaces an HTML artifact, accidental invocation can unnecessarily expose aggregated private financial data and create files the user did not intend to generate.

Vague Triggers

High

• Confidence: 97% confidence
• Finding: Deletion is a destructive action, yet the documented triggers include ambiguous everyday phrases like '撤销' and '删掉第5条'. In a conversational agent, ambiguous deletion intents can cause unintended loss of financial records, especially if the skill acts without a strong confirmation step tied to the exact record ID.

VirusTotal

64/64 vendors flagged this skill as clean.