Back to skill

Security audit

arena-power-user-playbook

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent offline helper for arena.ai workflows, with local file reads and writes that are disclosed and mostly user-directed.

Before installing, be aware that the README install command executes the latest ClawHub installer from npm, so a pinned or verified installer is safer when available. When using the skill, only pass files you are comfortable having processed locally, and choose explicit paths for state and log files.

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
README.md:25
Finding
Unpinned Remote Package Execution in Installation Instructions## Vulnerability Details **File Location**: `README.md`, line 25 **Vulnerability Type**: Supply-chain risk caused by executing a mutable package release **Risk Level**: Medium ### Vulnerable Code ```bash npx --yes clawhub@latest install @orionshaowswmw/arena-power-user-playbook ``` ### Technical Analysis The documented installation command directs `npx` to download and execute the mutable `latest` release of the `clawhub` package. The `--yes` option suppresses the normal installation confirmation, while neither the installer nor the requested Skill is pinned to an immutable version or verified digest before execution. Consequently, the code executed by this command may differ from the code reviewed during this audit. The risk is confined to users who follow this installation instruction; the bundled Python scripts themselves do not retrieve or execute remote payloads. ### Attack Path 1. An attacker compromises the package registry, the `clawhub` publisher account, or a dependency included in a future `latest` release. 2. The attacker publishes a modified release that contains malicious installation code. 3. A user follows the documented command. 4. `npx --yes` downloads and executes the attacker-controlled release without an interactive confirmation. 5. The malicious package runs with the privileges and environment access of the user who invoked the command. ### Impact Assessment Successful exploitation could allow arbitrary code execution under the installing user's account. Depending on that user's privileges and environment, the malicious installer could read accessible files and credentials, modify local data, install additional software, or establish persistence. The command does not itself request elevated privileges, so the direct privilege boundary is the invoking user's existing permissions.
Remediation
## Remediation Suggestions - Replace `clawhub@latest` with an exact, audited package version. - Pin the Skill to an exact version or immutable content digest where the registry supports it. - Remove `--yes` so unexpected package installation requires explicit confirmation. - Publish the expected installer and artifact hashes through a trusted channel. - Verify package provenance, signatures, and checksums before executing the installer. - Retain the existing post-installation tree-hash verification, but clarify that it does not protect against malicious code already executed by `npx`. - Prefer an installation workflow that downloads the package without executing it, verifies its integrity, and only then invokes reviewed code.
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Lp3

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding
The skill advertises and instructs use of local Python scripts that can read user-specified files, write local state/log files, and invoke a shell command interface, yet the skill declares no permissions. This creates a permission-transparency gap: users and hosting systems may not realize the skill can access or modify local files, increasing the risk of unintended data exposure or file tampering if the scripts are run in a sensitive workspace.

Static analysis

No suspicious patterns detected.