Security audit

Anthropic

Security checks for vulnerabilities and agentic risk

Overview

The skill is coherent for using Anthropic through OOMOL, but its first-time setup tells users to execute unverified remote installer scripts directly.

Review the installer path before installing. Prefer installing the oo CLI from a pinned, signed, or checksum-verified release channel, and do not run the provided remote script commands from an elevated shell. Once oo is already installed and connected, the skill's Anthropic connector usage is disclosed and reasonably scoped.

A.I.G

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 (2)

T03 · Remote Payload Retrieval and Execution

Error

Location: SKILL.md:58
Finding: Unverified Remote Shell Script Download and Execution
Content: View full analysis
Remediation: View remediation
/install.sh" echo " /tmp/oo-install.sh" | sha256sum --check bash /tmp/oo-install.sh ``` The temporary path should be created securely, and a verified package or signed binary is preferable to an installer script. ]]>

T03 · Remote Payload Retrieval and Execution

Error

Location: SKILL.md:62
Finding: Unverified Remote PowerShell Script Download and Execution
Content: View full analysis
Remediation: View remediation
/install.ps1" $path = Join-Path $env:TEMP "oo-install.ps1" Invoke-WebRequest -Uri $uri -OutFile $path if ((Get-FileHash $path -Algorithm SHA256).Hash -ne "") { throw "Installer integrity verification failed." } # Review and verify the publisher signature before execution. Get-AuthenticodeSignature $path & $path ``` Production instructions should reject invalid or unexpected signatures automatically rather than relying only on visual review. ]]>

SkillSpector

Vulnerability Patterns

Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (3)

External Script Fetching

High

Category: Supply Chain
Confidence: 98% confidence
Finding: The skill instructs users to install software via curl ... | bash, which executes a remote script directly without verification. If the hosting domain, transport path, or script content is compromised, this becomes arbitrary code execution on the user's machine; because the skill is framed as operational guidance, users may follow it with elevated trust.
Content: Scanner excerpt · SKILL.md (reported line 58)May include surrounding context.

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

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 96% confidence
Finding: The description says to use this skill for "ANY Anthropic request," including reading, creating, and updating data. That activation scope is extremely broad and lacks constraints or negative examples, which could cause the skill to be invoked for many loosely related mentions of Anthropic.
Content: No source excerpt is available for this finding.

Intent-Code Divergence

Medium

Category: Not specified by scanner
Confidence: 83% confidence
Finding: The file marks create_message with [write] and then states that [write] actions 'change Anthropic state'. But the listed Anthropic actions are count_message_tokens, create_message, get_model, and list_models; create_message is an inference call, not an operation that clearly creates or updates durable account data as the manifest description suggests. This documentation overstates the kind of side effect the action has and can mislead an agent about the real behavior.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.