Back to skill

Security audit

Media Studio Metered

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly coherent for metered media generation, but it asks users to install a mutable global CLI that handles credentials, billing, media uploads, and storage operations.

Review before installing. Prefer a pinned and verified settlemesh CLI version, use a restricted API key with spending limits, avoid exposing unrelated secrets in the environment, confirm costs before paid renders, and require explicit confirmation before deleting stored objects.

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:28
Finding

Unpinned Global Installation of a Security-Sensitive Third-Party CLI

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 28-30
Vulnerability Type: Insecure third-party dependency installation
Risk Level: Medium

Vulnerable Code

markdown
## Setup (once)

1. `npm install -g settlemesh@latest`
2. `settlemesh login` (a human approves once) — or set `SETTLE_API_KEY=sk-settle-...` for headless/CI.

The dependency is also introduced in the compatibility declaration at line 5:

yaml
compatibility: "Requires the settlemesh CLI (npm i -g settlemesh) and a SETTLE_API_KEY; makes authenticated, metered calls to the SettleMesh platform over the network."

Technical Analysis

The skill instructs the agent or user to install settlemesh@latest globally from npm. The latest tag is mutable and does not identify a reviewed, immutable package version. The project provides no lockfile, integrity hash, provenance verification, trusted-publisher verification, or upgrade-review process.

An npm package can execute lifecycle scripts during global installation and arbitrary code whenever its CLI is invoked. This CLI is particularly security-sensitive because the skill instructs it to:

  • Access SETTLE_API_KEY.
  • Cache an authenticated login session.
  • Upload locally selected media.
  • Submit prompts and media to remote services.
  • Make metered, potentially billable API requests.
  • Read and write generated media through object storage.

The finding does not establish that the current settlemesh package is malicious. The vulnerability is the unsafe dependency acquisition model, which permits the effective installed implementation to change after the skill has been reviewed.

Attack Path

  1. An attacker compromises the npm package, its publisher account, its release process, or the mutable release referenced by the latest tag.
  2. The attacker publishes a malicious version and assigns it to latest.
  3. Following the skill instructions, a user or agent r ...[truncated 1333 chars]
Remediation
View remediation

Remediation Suggestions

  1. Replace the mutable tag with an exact, reviewed version:
    bash
    npm install -g settlemesh@X.Y.Z
    
  2. Document the expected package name, publisher identity, source repository, and release provenance so users can detect package substitution or publisher compromise.
  3. Pin and verify package integrity using a trusted lockfile, expected registry integrity metadata, signed provenance, or a documented checksum where the installation workflow supports it.
  4. Avoid global installation where practical. Use a project-local dependency with a lockfile and invoke the pinned binary from node_modules/.bin, preferably inside a restricted container or dedicated low-privilege account.
  5. Disable npm lifecycle scripts during installation if the verified package does not require them:
    bash
    npm install --ignore-scripts --save-exact settlemesh@X.Y.Z
    
    This is defense in depth and does not replace source and provenance review.
  6. Establish an explicit upgrade process that reviews release changes before updating the pinned version.
  7. Give the API key only the minimum required permissions, apply spending limits where supported, rotate credentials after suspected compromise, and avoid exposing unrelated secrets in the CLI process environment.
  8. Clarify how cached login credentials are stored and protected, and provide logout, revocation, and secure-uninstallation procedures.
Vulnerability Patterns
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • 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
Findings (3)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
88% confidence
Finding

The documented settlemesh storage rm renders/clip1.mp4 operation is a destructive tool invocation and demonstrates removal of stored artifacts with no safeguards, confirmation step, or validation of user intent. In a skill that manages durable outputs for resale and storage, deletion is contextually more dangerous because it can remove customer deliverables or business assets.

Content

Scanner excerpt · SKILL.md (reported line 67)May include surrounding context.

settlemesh storage url renders/clip1.mp4 --ttl 3600 # presigned GET URL settlemesh storage ls renders/ --json settlemesh storage get renders/clip1.mp4 ./clip1.mp4 settlemesh storage rm renders/clip1.mp4

text

## Resell it (charge end users per render)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The manifest description says to use the skill when the user wants to "make an image," "generate video," "do text-to-speech," or "run a media/generation pipeline," which are broad intents likely to match many ordinary user requests. It does not provide narrower trigger constraints or exclusion conditions, increasing the risk of unintended invocation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The skill includes a destructive deletion command for object storage without any surrounding requirement for explicit user confirmation or warning about permanence and scope. In an agent context, users may ask to inspect or manage outputs and the agent could execute deletion as part of a workflow, causing unintended data loss.

Content

No source excerpt is available for this finding.