Back to skill

Security audit

MiniCut film studio

Security checks for vulnerabilities and agentic risk

Overview

The filmmaking workflow is plausible, but setup automatically executes unverified downloads and delegates further actions to instructions outside the reviewed package.

Review the installer and the full skill it installs before allowing automatic setup or command execution. Prefer a verified, fixed release and ordinary user permissions. Browser sign-in and paid media generation are expected, and the skill requires approval of each paid run. This review covers the bundled instructions, not the downloaded software or replacement skill.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

External Script Fetching

High
Category
Supply Chain
Confidence
99% confidence
Finding

The installation command downloads a mutable remote script and immediately executes it without review, version pinning, or independent integrity verification; the Windows alternative has the same weakness. If the distribution endpoint or release infrastructure is compromised, attacker-controlled code could run with the agent's local permissions and access credentials or modify files. Installing a CLI is legitimate in this context, but subsequent setup and delegation to a newly downloaded skill extend the trust placed in that distribution channel; there is no evidence here that the author is malicious.

Content

Scanner excerpt · SKILL.md (reported line 16)May include surrounding context.

md
## Step 0. Install, once per machine

1. Check: `minicut status`. If the command is found, go to Step 1.
2. Mac or Linux: `curl -fsSL https://minicut.ai/install.sh | bash -s -- --region global`
   Windows: `curl -fsSL https://minicut.ai/install.cmd -o install.cmd && install.cmd`
3. Run `MINICUT_FROM=<agent> minicut setup`, where `<agent>` is the tool you run in: `claude`,
   `codex`, `cursor`, `gemini` or `openclaw` (Windows: `set MINICUT_FROM=<agent>` first, then