T09 · Insecure Skill Coding Practices
- Location
scripts/meeting_claw.py:50- Finding
Arbitrary Local File Disclosure Through Missing Audio Validation
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
The skill is a plausible meeting-minutes tool, but it needs review because it can upload any readable file path to cloud services and can automatically delete more local workspace files than its documentation describes.
Review before installing or running this skill. Only use it with recordings you are authorized to send to Volcengine and OpenClaw, avoid passing arbitrary local paths, keep the Volcengine credentials narrowly scoped, and do not store unrelated files under the MeetingClaw workspace because old files there may be deleted automatically.
scripts/meeting_claw.py:50Arbitrary Local File Disclosure Through Missing Audio Validation
scripts/meeting_claw.py:119Overbroad Retention Cleanup Deletes Minutes and Unrelated Workspace Files
The document claims generated meeting notes are only stored locally and not uploaded to public servers, yet the workflow states that transcription text is sent to an external OpenClaw large model for analysis. For meeting content, transcripts often contain confidential business discussions and personal data, so this contradiction can mislead users into exposing sensitive information to a third party without informed consent.
The skill advertises capabilities that imply access to environment variables, local file read/write, and network services, but it does not declare any explicit tool scope or permissions boundary. This weakens reviewability and least-privilege enforcement, making it easier for a skill handling sensitive meeting audio and API credentials to access or transmit more than users expect.
The skill description does not clearly warn users that uploaded audio and derived transcript content are sent to external cloud providers for recognition and summarization. In a meeting-minutes context, users may reasonably assume local processing and unknowingly submit regulated, confidential, or privileged material to third parties.
The file title and all user-facing instructional content are written exclusively in Chinese, with no indication that users may choose another language or that the locale restriction is required for a region-specific purpose. Under the stated policy, forcing a specific language without opt-in is a natural-language policy concern.
The manifest describes uploading audio, transcribing it, and generating structured meeting minutes. While calling external speech services is consistent with that purpose, directly pulling long-lived credentials from environment variables introduces a broader capability to access host configuration secrets that is not stated or justified by the skill description.
The skill stores original meeting audio under a predictable local workspace path and retains it for 30 days, but this persistence is not disclosed in the description. Meeting audio commonly contains sensitive business or personal information, so undisclosed local retention increases the risk of privacy exposure, unauthorized local access, and accidental data accumulation.
The skill sends uploaded audio to an external speech service and then sends transcript content to an external LLM service without an explicit user-facing warning about third-party data transmission. Because meeting recordings often contain confidential discussions, personal data, or regulated information, silent exfiltration to external processors materially raises confidentiality and compliance risk.
The request parameters force the language to zh-CN, which imposes a specific locale setting on all users. The file does not offer language selection, opt-in, or any justification that this skill is intended only for Chinese-language audio.
The file presents the skill entirely in Chinese and describes Chinese-language outputs such as structured meeting minutes, but it does not say whether other languages are supported or whether the Chinese-only behavior is intentional and user-selectable. This can create a language-policy issue if the skill implicitly forces one language without opt-in.
The generated meeting minutes are written to disk even though the description only says they are generated, not persisted. While saving output is not inherently dangerous, meeting minutes often contain sensitive summaries, action items, and participant information, so undisclosed storage can create privacy and data-handling risk.
The cleanup routine deletes files older than the retention threshold from the workspace using os.remove. There is no user-facing notice at deletion time and no prior warning in this file that stored audio and generated artifacts will be automatically removed after 30 days.