Back to skill

Security audit

Pans Meeting Summarizer

Security checks for vulnerabilities and agentic risk

Overview

This skill is a straightforward meeting summarizer that handles sensitive meeting content through OpenAI APIs, so users should review privacy and output accuracy before use.

Install only if you are comfortable processing meeting audio or transcripts with OpenAI-backed tooling. Avoid using it on highly confidential customer calls unless your organization permits that data flow, and review generated summaries before relying on them or using any future CRM-sync feature.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/summarize.py:91
Finding

Prompt Injection Through Untrusted Meeting Transcript

Content
View full analysis
Dict[str, Any]: """ 生成结构化会议纪要 Args: transcript: 会议转录文本 context: 会议上下文/类型 Returns: 结构化纪要数据 """ if not self.client: # 无API时返回基础结构 return self._fallback_summary(transcript) prompt = f"""请分析以下会议记录,生成结构化的会议纪要。 会议类型:{context} 会议记录: {transcript} 请按以下JSON格式输出(不要添加markdown代码块标记): {{ "meeting_summary": {{ "topic": "会议主题", "date": "会议日期(如未提及则填今天)", "participants": ["参与人列表"], "key_points": ["讨论要点1", "讨论要点2"], "decisions": ["决策事项1", "决策事项2"], "next_steps": ["下一步计划"] }}, "action_items": [ {{ "task": "待办事项描述", "owner": "负责人", "priority": "high/medium/low", "due_date": "截止日期(如未提及则填待定)" }} ], "customer_insights": {{ "budget": "预算情况分析", "decision_makers": ["决策链关键人物"], "competitors": ["竞品情况"], "pain_points": ["痛点需求"], "buying_signals": ["购买信号"], "concerns": ["顾虑点"] }}, "follow_up_suggestions": ["后续跟进建议"] }} """ try: response = self.client.chat.completions.create( model="gpt-4o-mini", messages=[ {"role": "system", "content": "你是一个专业的销售会议分析师,擅长提取关键信息和洞察。"}, {"role": "user", "content": prompt} ], temperature=0.3, response_format={"type": "json_object"} ) ``` ### Technical Analysis The `transcript` and `context` values are treated as untrusted input, but they are interpolated directly into the same user message that contains the summarization instructions and required output schema. The applicatio ...[truncated 2601 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (8)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
83% confidence
Finding

The skill advertises capabilities that imply reading input files, writing outputs, and potentially using environment-provided credentials, but it does not declare any explicit tool scope or permissions. In a skill that can process local transcripts/audio and sync to CRM, this ambiguity increases the chance of unintended file access or unauthorized use of secrets/API credentials without clear user or platform controls.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The trigger phrases are broad and match many ordinary meeting-related requests, which can cause the skill to activate in contexts the user did not intend. Because the skill can process sensitive meeting content and may update CRM records, overbroad activation raises the risk of unnecessary data handling or accidental external actions.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The markdown trigger section repeats broad activation terms without qualifiers, guardrails, or exclusions, making accidental invocation more likely. In this context, an unintended match could expose sensitive meeting transcripts to summarization pipelines or prepare downstream CRM updates based on the wrong conversation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill supports CRM synchronization but does not warn users that it can modify external customer records. This omission is dangerous because meeting summaries and extracted action items may be inaccurate or incomplete, leading to accidental corruption of CRM data, disclosure to the wrong account, or unauthorized updates to business-critical records.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill states that it uses external AI services for transcription and summarization but does not disclose that meeting audio and transcripts may be transmitted off-platform. This is especially risky because sales meetings often contain confidential customer data, pricing, budgets, strategy, and personal information, creating material privacy, confidentiality, and compliance exposure.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

Natural-language strings, prompts, help text, and output instructions are written in Chinese, and the model prompt explicitly requests Chinese JSON field content such as meeting summaries and follow-up suggestions. There is no indication that users can choose another language or opt into this locale behavior.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill sends full meeting transcript content to an external OpenAI API, and meeting notes commonly contain sensitive business, customer, pricing, or personal data. Because the CLI does not provide a clear user-facing warning or explicit consent gate before transmission, users may unknowingly expose confidential information to a third-party service.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

The manifest describes a tool for turning meeting audio or text into summaries, action items, insights, and optionally syncing CRM. While using an API-backed model can support summarization, directly sourcing credentials from process environment adds a sensitive capability not declared in the manifest text and is not necessary when an explicit --api-key argument already exists.

Content

No source excerpt is available for this finding.