ClawHub
Back to skill

Security audit

SenseAudio 会议助手 / SenseAudio Meeting Assistant

Security checks for vulnerabilities and agentic risk

Overview

This meeting-assistant skill is coherent, but it needs review because it can send sensitive live or recorded meeting audio to SenseAudio without clear consent, retention, or access-control guidance.

Install only if you are comfortable sending meeting audio and transcripts to SenseAudio. Before use, confirm participant consent and organizational approval, avoid regulated or highly confidential meetings unless the service is sanctioned, keep the API key in environment or secret storage, restrict access to historical transcript retrieval, and define how raw recordings and logs are encrypted, retained, and deleted.

A.I.G

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (14)

Natural-Language Policy Violations

Medium
Confidence
92% confidence
Finding
The title and core usage text are presented bilingually, and the skill consistently frames outputs and interaction structure in Chinese/English form rather than offering a user language choice. This can violate language/locale policy when a skill imposes a specific language format without explicit opt-in or documented regional justification.

Vague Triggers

Medium
Confidence
92% confidence
Finding
The skill enables implicit invocation (`allow_implicit_invocation: true`) without any visible trigger constraints or narrowing conditions in this file. That can cause the assistant to activate in broader contexts than intended, increasing the chance of accidental invocation, unintended access to meeting-related behaviors, or prompt-routing abuse if a user message loosely matches the skill’s purpose.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The examples show sending live meeting audio and generating transcripts/notes using an API key, but they omit any warning about consent, sensitive-content handling, storage, or secure credential management. In a meeting-assistant context, this can lead users to transmit confidential conversations or mishandle API secrets without understanding the privacy and security implications.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The recommended workflow explicitly advises saving raw meeting audio, but provides no guidance on retention limits, access controls, encryption, or deletion. Because raw recordings often contain the fullest copy of sensitive meeting content, this omission increases the risk of unnecessary retention, unauthorized access, and compliance violations.

External Transmission

Medium
Category
Data Exfiltration
Content
If recognition quality is poor, analyze the file first.

```bash
curl -X POST https://api.senseaudio.cn/v1/audio/analysis \
  -H "Authorization: Bearer $SENSEAUDIO_API_KEY" \
  -F "model=sense-asr-check" \
  -F "file=@meeting.wav"
Confidence
50% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
If recognition quality is poor, analyze the file first.

```bash
curl -X POST https://api.senseaudio.cn/v1/audio/analysis \
  -H "Authorization: Bearer $SENSEAUDIO_API_KEY" \
  -F "model=sense-asr-check" \
  -F "file=@meeting.wav"
Confidence
50% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
# 离线会议转写 / Offline Meeting Transcription

当用户提供会议录音或会后文件时，使用 `POST https://api.senseaudio.cn/v1/audio/transcriptions`。
Use `POST https://api.senseaudio.cn/v1/audio/transcriptions` for uploaded meeting recordings.

## 适用场景 / Best Fit
Confidence
89% confidence
Finding
This line directs users to send uploaded meeting recordings to an external API endpoint, which creates a real data exfiltration path for sensitive audio content. In a meeting-assistant context, the transmitted files may contain confidential, regulated, or personal information, making external transmission materially risky unless clearly disclosed and controlled.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The skill explicitly instructs uploading meeting recordings to a remote third-party API but does not include any user-facing warning, consent step, or privacy guidance for potentially sensitive meeting content. Because meeting audio commonly contains confidential business discussions or personal data, this omission can lead to unintended external disclosure even if the transmission itself is technically expected for the feature.

External Transmission

Medium
Category
Data Exfiltration
Content
# 离线会议转写 / Offline Meeting Transcription

当用户提供会议录音或会后文件时，使用 `POST https://api.senseaudio.cn/v1/audio/transcriptions`。
Use `POST https://api.senseaudio.cn/v1/audio/transcriptions` for uploaded meeting recordings.

## 适用场景 / Best Fit
Confidence
89% confidence
Finding
The English instruction duplicates the external upload behavior, increasing the likelihood that users will transmit meeting recordings off-platform without recognizing the privacy implications. The risk is not the presence of a URL alone, but the operational instruction to upload potentially sensitive meeting data to a remote service.

External Transmission

Medium
Category
Data Exfiltration
Content
## 接口信息 / Endpoint

- 地址 / URL：`https://api.senseaudio.cn/v1/audio/transcriptions`
- 方法 / Method：`POST`
- Content-Type：`multipart/form-data`
- 鉴权 / Auth：`Authorization: Bearer <SENSEAUDIO_API_KEY>`
Confidence
84% confidence
Finding
Documenting the endpoint and bearer-token authentication formalizes a workflow for transmitting raw meeting media to an external service. In this context, that is security-relevant because it enables offsite transfer of sensitive recordings, and the file provides no caution around data sensitivity, approval requirements, or secure operational handling.

External Transmission

Medium
Category
Data Exfiltration
Content
## 会议录音推荐请求 / Recommended Request for Meetings

```bash
curl https://api.senseaudio.cn/v1/audio/transcriptions \
  -H "Authorization: Bearer $SENSEAUDIO_API_KEY" \
  -F "file=@meeting.wav" \
  -F "model=sense-asr-pro" \
Confidence
90% confidence
Finding
The sample curl command provides a ready-to-run example for uploading a meeting recording file to a remote transcription service, making external disclosure easy to perform. Given the skill’s purpose, this is contextually more dangerous because users may paste in real meeting audio containing sensitive discussion, identities, and action items.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The workflow explicitly instructs sending live meeting audio, and later recommends logging identifiers like session_id and trace_id, but it does not mention obtaining participant consent, disclosing third-party transmission, or applying privacy controls. In a meeting-assistant context, this can expose sensitive conversations and metadata to an external service without adequate user awareness, creating privacy, compliance, and trust risks.

Missing User Warnings

Low
Confidence
88% confidence
Finding
The historical transcript retrieval example accesses prior meeting records, which are likely sensitive, but it lacks any warning about authorization, confidentiality, or safe handling of past transcripts. This may normalize broad retrieval of archived meeting data without verifying that the requester is entitled to access it.

Natural-Language Policy Violations

Low
Confidence
87% confidence
Finding
The example sets `target_language` to `en`, which can steer implementations toward forcing English output by default. Because the document does not frame this as a user-selectable example or justify an English-only locale constraint, it risks violating language/locale choice expectations.

Static analysis

No suspicious patterns detected.