Back to skill

Security audit

Replicate

Security checks for vulnerabilities and agentic risk

Overview

The skill is purpose-aligned for Replicate, but its setup instructions ask users to run an unverified remote installer directly in a shell.

Review the setup path before installing. Prefer installing the oo CLI through a verified, versioned release or package manager, and do not run the curl-to-bash or PowerShell iex command unless you trust OOMOL's installer endpoint and accept that it can execute local code. For normal Replicate use, confirm any create or cancel action payload before it runs.

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:62
Finding
Unverified Remote Installer Downloaded and Executed Directly## Vulnerability Details **File Location**: `SKILL.md`, lines 62–70 **Vulnerability Type**: Remote payload retrieval and execution **Risk Level**: Critical The first-time setup instructions execute remotely hosted installation scripts directly through Bash and PowerShell: ```markdown - **`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 irm https://cli.oomol.com/install.ps1 | iex # Windows PowerShell ``` ``` ### Technical Analysis Both installation commands retrieve mutable code from an external URL and immediately pass it to a command interpreter. The instructions do not pin an installer version, verify a cryptographic signature or checksum, or provide an opportunity to inspect the downloaded content before execution. HTTPS protects data in transit but does not establish that the current payload is identical to the one reviewed with this Skill. Compromise of the hosting service, publishing account, DNS infrastructure, TLS termination environment, or release pipeline could cause arbitrary replacement code to be executed. Because the effective installer can change independently of the audited repository, the package does not fully define the code users are instructed to run. Installing the required CLI is relevant to the declared functionality, but direct, unverified pipe-to-shell execution exceeds the minimum privileges and trust required. A version-pinned package or a separately downloaded and cryptographically verified installer would provide the same functionality with less supply-chain exposure. The audit found no evidence establishing that the current remote scripts are malicious. The vulnerability is the unsafe execution mechanism and its dependency on mutable, unaudited remote content. ### Attack Path 1. The `oo` CLI is absent, and an ...[truncated 1421 chars]
Remediation
## Remediation Suggestions 1. Remove both direct pipe-to-interpreter installation commands. 2. Pin the CLI to a specific audited release instead of retrieving a mutable generic installer. 3. Download the installer or binary as a separate file before execution. 4. Verify it against a vendor-published cryptographic signature or a checksum obtained through an independently protected channel. 5. Abort installation if verification fails; do not offer an automatic bypass. 6. Present the verified file, source, version, and expected installation effects to the user and obtain explicit approval before execution. 7. Prefer a reputable platform package manager with repository signing and version pinning where available. 8. Avoid requiring administrative privileges unless a documented installation step strictly needs them. 9. Publish reproducible release artifacts and retain immutable, versioned download URLs so the reviewed payload can be independently validated. 10. If installer scripts remain necessary, document their required filesystem, process, and network permissions and ensure they operate with least privilege.
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
97% confidence
Finding
The skill instructs the agent to install software by piping a remotely fetched script directly into a shell (`curl ... | bash`). This creates a supply-chain and remote-code-execution risk: if the server, transport, distribution path, or script content is compromised, arbitrary code would run immediately on the host without inspection or integrity verification.

Vague Triggers

Medium
Confidence
96% confidence
Finding
The description says to use this skill for "ANY Replicate request" and "Whenever a task involves Replicate," which is extremely broad and lacks boundary conditions. This can cause unintended invocation for casual mentions or tasks only tangentially related to Replicate because no negative examples or narrower trigger constraints are provided.

Static analysis

No suspicious patterns detected.