Security audit

跨会议知识库

Security checks for vulnerabilities and agentic risk

Overview

The skill’s meeting knowledge-base purpose is coherent, but it can aggregate sensitive meeting records from Tencent Meeting or local notes into a persistent index from broad prompts without clear scope confirmation.

Install only if you are comfortable letting the agent search and summarize meeting records. Before use, explicitly limit the meeting date range, topics, participants, or local folder, and review the generated knowledge-base file because it may contain internal decisions, action items, risks, and names.

SkillSpector

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The trigger phrases are broad consumer-style requests such as building a knowledge base, checking prior decisions, or summarizing recent meeting conclusions. In an agent environment, this can cause the skill to activate on ordinary conversation and automatically aggregate sensitive meeting records from Tencent Meeting or local notes without the user clearly intending data retrieval at that moment.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The README encourages natural-language use but does not prominently warn that the skill may pull and aggregate records from Tencent Meeting or read local meeting-note directories. This creates a transparency and consent problem: users may issue innocuous-looking requests like '帮我把最近的会议结论汇总一下' without realizing the skill could process a broad set of potentially confidential internal records.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The skill explicitly says users can describe needs in everyday language and that the AI will automatically recognize and invoke the skill, which creates overly broad activation conditions. This can cause the skill to trigger on ambiguous requests and access or process meeting-derived data when the user did not clearly intend to invoke this specific workflow.

Vague Triggers

Medium

Confidence: 86% confidence
Finding: The listed trigger phrases are common conversational requests such as asking what was decided previously or what tasks remain, and they lack strong disambiguation. In a context involving internal meeting records, broad triggers increase the risk of unintentional activation and unintended disclosure or processing of sensitive organizational information.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The trigger examples are broad natural-language phrases such as '帮我建一个会议知识库', '之前会议怎么定的', and '目前项目有什么风险', which can overlap with ordinary conversation in many enterprise contexts. In an agent environment, this increases the chance of accidental skill invocation and unintended retrieval, aggregation, or disclosure of cross-meeting knowledge base content that may contain sensitive decisions, action items, or risks.

Static analysis

No suspicious patterns detected.