Back to skill

Security audit

StealthGPT

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly coherent, but its setup guidance includes direct execution of a remote installer, so users should review it before installing.

Install only if you are comfortable using OOMOL as the connector for your StealthGPT account. Do not blindly run the one-line installer commands; prefer official, version-pinned, signed, or checksum-verified installation steps, and review the oo CLI setup before connecting credentials or using paid StealthGPT actions.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:58
Finding
Unverified Remote Installer Download and Immediate Shell Execution## Vulnerability Details **File Location**: `SKILL.md`, lines 58–62 **Vulnerability Type**: Remote payload retrieval and execution **Risk Level**: Critical ### Vulnerable Code ```bash curl -fsSL https://cli.oomol.com/install.sh | bash # macOS / Linux ``` ```powershell irm https://cli.oomol.com/install.ps1 | iex # Windows PowerShell ``` ### Technical Analysis The installation instructions retrieve mutable scripts from `cli.oomol.com` and immediately execute them through Bash or PowerShell. The downloaded content is not pinned to a specific release, saved for inspection, authenticated with a publisher signature, or checked against a known cryptographic hash. HTTPS provides transport protection but does not establish that the script remains unchanged after the Skill has been audited. A compromised hosting account, web server, release process, or domain could replace the installer with arbitrary code. The effective executable payload is therefore controlled by remote infrastructure at invocation time and cannot be determined from the reviewed project. The installation behavior also exceeds the minimum runtime privileges needed to invoke the declared StealthGPT connector. Although installing the required CLI may be necessary during initial setup, executing an unverified remote script is not necessary; a pinned and independently verified package can provide the same functionality. ### Attack Path 1. The `oo` command is unavailable, causing the Agent or user to follow the first-time setup instructions. 2. The shell retrieves the current contents of `install.sh` or `install.ps1` from the remote server. 3. The response is piped directly into Bash or `Invoke-Expression`, without local review or integrity verification. 4. If the distribution infrastructure or installer has been maliciously modified, attacker-controlled commands execute with the privileges of the invoking user. 5. The payload can access user-readable data, modify user-owned files an ...[truncated 835 chars]
Remediation
## Remediation Suggestions 1. Remove both direct execution pipelines (`curl | bash` and `irm | iex`). 2. Pin installation instructions to a specific, immutable CLI release and version. 3. Download the installer or package to a local file without executing it automatically. 4. Publish and require verification of a hard-coded SHA-256 or stronger checksum obtained through an independently protected channel. 5. Prefer signed platform packages and verify the publisher signature before installation. 6. Allow the user to inspect the downloaded artifact and require explicit approval before execution. 7. Run installation with ordinary user privileges unless a documented operation strictly requires elevation. 8. Document the files, permissions, subprocesses, and network endpoints used by the installer. 9. For automated Agent workflows, stop when the CLI is missing and provide safe manual installation guidance rather than executing a remote installer. 10. Consider distributing the CLI through reputable, version-pinned package managers with signature and provenance verification.
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • 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 (2)

External Script Fetching

High
Category
Supply Chain
Content
- **`oo: command not found`** — install the oo CLI (other platforms: <https://cli.oomol.com/install-guide.md>):

  ```bash
  curl -fsSL https://cli.oomol.com/install.sh | bash    # macOS / Linux
  ```

  ```powershell
Confidence
96% confidence
Finding
The skill instructs the agent to install software by piping a remotely fetched script directly into a shell (`curl ... | bash`), which is a classic arbitrary code execution risk. If the remote host, CDN, DNS, TLS termination, or installation script is compromised, an agent or user following these instructions could execute attacker-controlled code on the local system.

Vague Triggers

Medium
Confidence
94% confidence
Finding
The invocation text says to use this skill for "ANY StealthGPT request" and "instead of calling the API directly," which is overly broad and can cause the agent to invoke the skill in situations where it is not necessary or where safer/native handling would be preferable. Broad routing language increases the chance of unintended tool use and expands the blast radius if the skill later gains risky actions or misconfigured permissions.

Static analysis

No suspicious patterns detected.