Back to skill

Security audit

口播视频加工

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this video overlay skill suspicious because scanners report several potential data-loss and privacy concerns, though they do not report clear malicious behavior; coverage is partial.

Before installing, consider whether the workflow's deletion steps could affect files you want to keep, and whether you are comfortable with the cloud fallback's reported privacy gap. The scanner coverage is marked partial because some path references could not be resolved, so the available results do not establish that all behavior was assessed.

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
Vulnerability Patterns
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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)

The workflow instructs force-deleting an existing staged video without warning about potential data loss.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The workaround directs deletion of the existing input video before verifying that a replacement can be copied successfully. Although this is a staging path rather than an arbitrary user file, it may contain the only remaining copy or prior work, making unconditional deletion a data-loss risk.

Content

No source excerpt is available for this finding.

The cloud-GPU fallback lacks a privacy warning about sending identifiable face footage and audio to third-party infrastructure.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The fallback recommends third-party cloud processing of identifiable face footage and potentially voice audio without requiring approval for that transfer. Unlike the advertised local workflow, this crosses a privacy boundary and could expose sensitive media through provider access, storage, or retention; the separate likeness-authorization note does not address cloud disclosure.

Content

No source excerpt is available for this finding.

The read-only-file workaround instructs the agent to force-delete the existing source video without warning about data loss.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The proposed read-only workaround deletes the existing video before a replacement has been successfully staged. The sample-copy context reduces the expected risk, but the instructions do not verify that the existing file is disposable or backed up, so a failed copy can leave the project without its input.

Content

No source excerpt is available for this finding.

Existing PNG files in the work frame directories are silently deleted before rendering.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The script deletes every PNG in work/frames1, work/frames2, and work/frames3 without confirmation or a comment or message disclosing the cleanup. The introductory comments describe rendering and compositing, but do not explain that existing files in these directories will be removed.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx remotion'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.