Back to skill

Security audit

Industrial Mecha Style Transform (Seedance 2.0)

Security checks for vulnerabilities and agentic risk

Overview

The skill is a disclosed WeryAI video generator, but its bundled CLI can upload local files and its documented shell examples create command-injection risk.

Install only if you are comfortable giving this skill a WeryAI API key and outbound network access. Avoid local image paths entirely, use only public HTTPS image URLs, and run it in an isolated environment. Agents should invoke the Node script with structured arguments rather than constructing shell command strings from user prompts.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:135
Finding

Shell command injection through unescaped prompt interpolation

Content
View full analysis
","duration":10,"aspect_ratio":"9:16","resolution":"720p","generate_audio":true}' ``` The API documentation similarly instructs callers to place prompt data directly inside a shell command: ```sh node {baseDir}/scripts/video_gen.js wait --json '{"model":"KLING_V3_0_PRO","prompt":"...","duration":5,"aspect_ratio":"9:16"}' node {baseDir}/scripts/video_gen.js wait --json '{"model":"...","prompt":"...","image":"https://...","duration":5}' node {baseDir}/scripts/video_gen.js wait --json '{"model":"...","prompt":"...","image":"./girl.png","duration":5}' node {baseDir}/scripts/video_gen.js wait --json '{"model":"...","prompt":"...","images":["https://..."],"duration":5}' node {baseDir}/scripts/video_gen.js wait --json '{"model":"...","prompt":"...","images":["./frame-1.png","./frame-2.png"],"duration":5}' node {baseDir}/scripts/video_gen.js wait --json '...' --dry-run ``` ### Technical Analysis The Skill requires an agent to expand user-controlled text into a production prompt and then execute a documented command in which the JSON is enclosed by single shell quotes. No instruction requires shell-safe encoding, and the CLI offers no stdin-based JSON input. JSON escaping does not provide shell escaping. If the generated prompt contains a single quote, that character terminates the shell-quoted argument. Subsequent prompt content can then be interpreted as shell operators or commands. The JavaScript program itself parses `process.argv` without invoking a shell. The vulnerability arises when an agent follows the documented invocation by constructing a command string and executing it through a shell. Conseque ...[truncated 1198 chars]
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
scripts/video_gen.js:204
Finding

Unrestricted local-file read and upload to a third-party service

Content
View full analysis
controller.abort(), 60_000); let res; try { res = await fetch(MODELS_BASE_URL + UPLOAD_API_PATH, { method: 'POST', headers: { Authorization: `Bearer ${apiKey}`, }, body: form, signal: controller.signal, }); } catch (err) { clearTimeout(timer); if (err.name === 'AbortError') { throw new Error(`Upload timeout: ${filePath}`); } throw err; } clearTimeout(timer); let data; try { data = await res.json(); } catch { throw new Error(`Upload failed with non-JSON response (HTTP ${res.status}).`); } const wrapped = { httpStatus: res.status, ...data }; if (!isApiSuccess(wrapped)) { const apiErr = fo ...[truncated 3215 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (12)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill text claims tight restrictions to SEEDANCE_2_0 and discourages local file use, but it also admits the underlying script may support arbitrary models, model enumeration, local image reads, uploads, and broader video generation workflows. This mismatch is dangerous because reviewers and users may trust the narrower description while the actual executable behavior can exfiltrate local images, incur unexpected paid API calls, or operate outside the stated scope.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 20)May include surrounding context.

md
- `WERYAI_API_KEY` **must be set** before running `video_gen.js`.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 21)May include surrounding context.

md
- `WERYAI_API_KEY` **must be set** before running `video_gen.js`.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 22)May include surrounding context.

md
- `WERYAI_API_KEY` **must be set** before running `video_gen.js`.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 28)May include surrounding context.

md
- `WERYAI_API_KEY` **must be set** before running `video_gen.js`.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 29)May include surrounding context.

md
- `WERYAI_API_KEY` **must be set** before running `video_gen.js`.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 30)May include surrounding context.

md
- `WERYAI_API_KEY` **must be set** before running `video_gen.js`.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 34)May include surrounding context.

md
- `WERYAI_API_KEY` **must be set** before running `video_gen.js`.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
94% confidence
Finding

The skill declares network and environment-secret dependencies but does not define an explicit tool scope such as allowed tools or permissions. That weakens least-privilege controls and makes it easier for an agent runtime to grant broader execution or network access than users expect, especially since the skill invokes Node scripts and paid external APIs.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The instructions explicitly require converting prompts into a full English production prompt. This forces a specific language for operation without indicating that users may choose another language or opt in, which is a natural-language locale policy issue under the stated rules.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The trigger test prompt is broad enough to match general 'factory mecha' styling requests without explicitly requiring the skill name or tighter activation boundaries. This can cause unintended skill activation on loosely related user inputs, leading to incorrect routing, over-triggering, and application of this skill in contexts where it was not requested.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This documentation explicitly states that local image paths and file:// sources are automatically uploaded to a public URL before generation, but it does not provide a clear privacy or data-disclosure warning to users. That creates a real risk that operators may supply sensitive local files under the assumption they remain local, causing unintended exposure of proprietary, personal, or regulated image data to a third-party service.

Content

No source excerpt is available for this finding.