uni-vision-engine

AdvisoryAudited by Static analysis on May 10, 2026.

Overview

Detected: suspicious.dangerous_exec

Findings (1)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

ConcernHigh Confidence

ASI03: Identity and Privilege Abuse

What this means

The agent may use a preloaded account session and spend credits or act under an account identity that the user did not explicitly configure.

Why it was flagged

The script includes a default cached authentication session and later sends it as a Bearer token, so the skill can use account authority even when the user does not provide their own session.

Skill content

let sessionToken = "b79fc...39978"; // 默认缓存的认证 session

Recommendation

Remove the embedded session token, require the user to provide their own credential through a declared secure mechanism, and clearly document credit-spending behavior.

ConcernHigh Confidence

ASI05: Unexpected Code Execution

What this means

Running the skill can unexpectedly fetch and execute package-install behavior from npm on the user's machine.

Why it was flagged

If the dependency is missing, the script automatically runs a shell-based npm install during execution, downloading unpinned code outside the declared install process.

Skill content

const { execSync } = require('child_process');
execSync('npm install form-data --no-save', { stdio: 'ignore' });

Recommendation

Move dependency installation into a reviewed install spec, pin package versions with a lockfile, and avoid runtime npm installs.

NoteHigh Confidence

ASI02: Tool Misuse and Exploitation

What this means

When invoked, the agent has broad local shell capability rather than only a narrowly limited video-generation action.

Why it was flagged

The skill grants broad Bash access. This is used for the stated local video-generation workflow, but it is broader than a single scoped command.

Skill content

allowed-tools: Bash(*)

Recommendation

Prefer limiting allowed commands to the specific script and required log-reading commands, or ask for user confirmation before shell actions.

NoteHigh Confidence

ASI07: Insecure Inter-Agent Communication

What this means

Images provided in chat may be saved locally and submitted to the local Jimeng service and its generation/moderation flow.

Why it was flagged

The skill is designed to extract user-sent images and send them into the video-generation pipeline. That is purpose-aligned, but it is sensitive media handling.

Skill content

automatically intercept clothing/character images sent by the user in the chat interface and seamlessly pass them to the generation model

Recommendation

Use only with images the user intends to process, disclose the upload/moderation behavior, and clean up temporary image files after generation.

Findings (1)

critical

suspicious.dangerous_exec

Location: scripts/generate.js:9
Finding: Shell command execution detected (child_process).