Back to skill

Security audit

caption-styles

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this caption-style skill benign because its reported purpose and behavior appear aligned, while noting that its instructions use an unpinned HyperFrames command; confidence is medium and coverage is incomplete.

The verdict is benign, not risk-free. Before installing, consider whether you are comfortable with instructions that run an unpinned HyperFrames package; use a reviewed, fixed version if you follow them. Scanner coverage was partial: 41 of 42 components were scanned, with one partly inspected and a static analysis limit, so the evidence is incomplete.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

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

Medium
Category
MCP Rug Pull
Confidence
90% confidence
Finding

The documented npx command lacks an exact package version and can fetch and execute a changing registry release when the package is not installed locally. Although HyperFrames is a legitimate rendering CLI rather than an identified MCP server, a compromised release could execute with the agent's permissions; the surrounding reference to version 0.8 does not pin the command.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The catalog lookup uses an unversioned npx invocation, which can download and execute a mutable package release. A read-oriented catalog operation still executes the CLI locally, so a compromised release could access project files or credentials despite the benign caption-selection purpose.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The build step explicitly instructs execution of an unpinned npm package through npx. If registry resolution supplies a compromised release, that code executes with the agent's permissions; its legitimate role in adding caption components does not remove this supply-chain exposure.

Content

No source excerpt is available for this finding.

The build workflow instructs package execution without warning about third-party code execution or project changes.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

L38 directs the agent to run npx hyperframes add caption-highlight, rather than merely listing an available command. This can download and execute third-party package code and add component files to the project, but the document gives no warning about these effects.

Content

No source excerpt is available for this finding.

The transcription workflow offers a hosted service without warning that user audio would be uploaded.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

L41 instructs obtaining word timestamps using WhisperX or ElevenLabs Speech to Text. Choosing ElevenLabs entails transmitting the user's audio to a third-party service, potentially exposing private speech, but the document does not disclose that transfer or require user approval.

Content

No source excerpt is available for this finding.