Back to skill

Security audit

varg-previz

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because it reports a possible project-file exposure flaw and unsafe handling of storyboard content, alongside a workflow that may upload project media; some analysis was incomplete.

Consider the suspicious verdict before installing, especially if project files or media are sensitive. Review whether uploads to the external service are needed and authorized, and treat the reported project-file exposure and storyboard-content risks as unresolved. Skillspector inspected 19 of 27 components fully, with 8 partially inspected, so the scan evidence is incomplete.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
templates/previz/serve.py
Finding

Missing Host validation on GET requests permits DNS-rebinding access to project files

Content
View full analysis

Impact

Unauthorized disclosure of project content, including storyboards, production notes, media, and other readable files under the served project root. This finding does not establish access outside that root.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (6)

Code capability 'network' detected in templates/previz/js/main.js but not covered by declared permissions.

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'network' capability that is not covered by the declared 'allowed-tools' preapproval. Review the host runtime's actual enforcement policy; runtime permissions are unknown unless separately verified, and this declaration alone does not establish a permission ceiling.

Content

No source excerpt is available for this finding.

Code capability 'env' detected in templates/previz/seedance/tools/capture.mjs but not covered by declared permissions.

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'env' capability that is not covered by the declared 'allowed-tools' preapproval. Review the host runtime's actual enforcement policy; runtime permissions are unknown unless separately verified, and this declaration alone does not establish a permission ceiling.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/seedance-v2v.md (reported line 76)May include surrounding context.

md
## Run it

1. `check_balance` and `get_current_team` (the connection may bill a team; switching is the user's call).
2. Per file: `create_upload` (with `sha256:` hash for dedup) → `curl -X PUT -H "Content-Type: application/octet-stream"
   --data-binary @file "https://api.varg.ai<proxy_url>"` (auto-registers). Paste URLs into `pack.json` refs (`url`)
   and `requests.json`.
3. `estimate_cost` on one request, quote the total as a ceiling, get an explicit go. First time on a project: run ONE

Project media is uploaded to a third-party service without a privacy warning or prior user approval.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

The workflow uploads project media before obtaining the explicit go-ahead, which appears only in the subsequent cost-estimation step. Although uploading is expected for varg generation, a user requesting local previz or pack preparation could have confidential media transmitted before authorizing external processing.

Content

No source excerpt is available for this finding.

Storyboard-derived subject names reach innerHTML without escaping, contradicting the stated escaping policy.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

A storyboard-controlled sh.ref value is interpolated into refOf's descriptive label, propagated through describe into d.size, and inserted unescaped into #p-spec.innerHTML. A malicious subject name containing HTML with an event handler can execute script in the previz origin when the shot is displayed, potentially accessing project resources and same-origin save/export APIs.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · templates/previz/seedance/tools/build_pack.py (reported line 108)May include surrounding context.

python
def run(*a):
    subprocess.run(a, check=True)


def build(frames: Path):