Back to skill

Security audit

Seedance 2.5 Image to Video — 720p Still-to-Video with Native Audio

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed RunComfy video-generation skill that sends a user-provided prompt and public image URL to a paid third-party model, with some install and trigger wording users should review.

Install only if you intend to use RunComfy for paid image-to-video generation and are comfortable sending the prompt and public image URL to that service. Prefer a pinned or isolated CLI install when possible, avoid private-token URLs, and confirm provider/cost before acting on generic requests like "animate this image."

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:50
Finding

Unpinned Global Installation of a Third-Party CLI

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:50
Vulnerability Type: Unpinned third-party dependency installed globally
Risk Level: Medium

Complete Code Snippet

markdown
1. **RunComfy CLI** — `npm i -g @runcomfy/cli`

Technical Analysis

The prerequisite installs the latest available release of @runcomfy/cli globally without pinning a reviewed version or verifying its integrity. Because the npm package reference is mutable, the code installed when a user follows this instruction may differ from the version that existed when the Skill was audited.

npm installation may execute package lifecycle scripts, and the installed CLI later processes authentication tokens, prompts, image URLs, output files, and network responses. A compromise of the package, its publisher account, or its dependency chain could therefore introduce arbitrary local behavior. The global installation scope also makes the CLI available outside this Skill and increases the duration and breadth of exposure.

This does not establish that the current package is malicious. The vulnerability is the unsafe, unpinned supply-chain trust model prescribed by the Skill.

Attack Path

  1. An attacker compromises the @runcomfy/cli package, its publisher account, or a transitive dependency used by a future release.
  2. The attacker publishes a malicious package version under the legitimate package name.
  3. A user follows SKILL.md:50 and runs npm i -g @runcomfy/cli.
  4. npm resolves the mutable package reference to the compromised release.
  5. Malicious lifecycle scripts may execute during installation, or malicious code may execute when the CLI is invoked.
  6. The payload operates with the privileges of the installing or invoking user and may access data available to that process.

Impact Assessment

Successful exploitation could permit arbitrary code execution with the current user's privileges. Depending on the user's enviro ...[truncated 483 chars]

Remediation
View remediation

Remediation Suggestions

  • Pin the CLI to a specifically reviewed version, for example @runcomfy/cli@<approved-version>.
  • Prefer a project-local dependency governed by a committed lockfile instead of a global installation.
  • Use npm integrity and provenance verification controls and document the expected package publisher and registry.
  • Review the pinned package's lifecycle scripts and transitive dependencies before recommending installation.
  • Run the CLI under a minimally privileged account or isolated environment with access only to required inputs, outputs, and credentials.
  • Establish an explicit dependency-update process that requires security review before changing the pinned version.
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (1)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger list contains broad natural-language phrases such as "animate this image" and "still to video" that are likely to appear in ordinary user requests. This can cause unintended auto-invocation of the skill, potentially sending user-provided image URLs and prompts to the third-party RunComfy service without the user's explicit intent to use this specific provider or paid model.

Content

No source excerpt is available for this finding.