Back to skill

Security audit

Agent Token Sentinel

Security checks for vulnerabilities and agentic risk

Overview

The skill is a plausible cost-control helper, but it asks users to run a floating npx package for a powerful monitoring and process-killing role without a pinned or locally reviewable implementation.

Review this before installing. The concept matches a budget guard, but you should only run it in a least-privileged environment and prefer a pinned, audited OpenClaw package version with clear documentation of what it can read, what it can terminate, and how budget enforcement can be reversed or disabled.

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
SKILL.md:21
Finding
Unpinned Third-Party Package Execution Through npx## Vulnerability Details **File Location**: `SKILL.md`, lines 21–23 **Vulnerability Type**: Unpinned and unaudited third-party dependency execution **Risk Level**: Medium **Complete Code Snippet**: ```bash npx openclaw skill run agent-token-sentinel --cap 5.00 ``` ### Technical Analysis The documented usage instructs users to execute the `openclaw` npm package through `npx` without specifying an exact package version. Depending on the local environment and cache state, `npx` may retrieve the currently published package from the configured npm registry and execute its code. The project does not provide a lockfile, integrity hash, vendored implementation, trusted source reference, or other mechanism for verifying the package that will run. Consequently, the effective executable behavior is mutable and cannot be audited from the contents of this repository. The repository itself contains only `SKILL.md`; the advertised monitoring, loop detection, budget enforcement, and alert functionality is not implemented locally. Exploitation requires the package source or dependency chain to become malicious or resolve to an unintended artifact. Possible supply-chain scenarios include compromise of the publisher account, compromise of a transitive dependency, registry or configuration manipulation, or publication of an unexpectedly changed release. ### Attack Path 1. An attacker compromises the `openclaw` package, its publisher, one of its dependencies, or the registry resolution path used by the victim. 2. The attacker publishes or causes resolution to a malicious package version. 3. A user follows the usage instruction in `SKILL.md`. 4. `npx` resolves and downloads the unpinned package when an appropriate cached copy is unavailable. 5. The retrieved package code executes under the identity and permissions of the invoking user. 6. Malicious package code may access resources available to that user, subject to operating-system and enviro ...[truncated 536 chars]
Remediation
## Remediation Suggestions - Pin `openclaw` to a specific, audited version rather than allowing `npx` to select the current registry release. - Commit an appropriate lockfile containing resolved versions and integrity metadata. - Verify package provenance and integrity before execution, and document the expected publisher, registry, and source repository. - Avoid automatic execution of remotely resolved packages. Prefer installing reviewed dependencies through a controlled build process and invoking the verified local binary. - Consider vendoring or including the actual skill implementation so its behavior can be reviewed with the repository. - Use a trusted private registry or dependency proxy with package allowlisting, malware scanning, and immutable artifacts. - Run the skill in a least-privileged, isolated environment without unnecessary credentials, sensitive filesystem access, or unrestricted network access. - Add automated dependency monitoring and require security review before updating the pinned version.
Vulnerability Patterns
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding
The skill instructs users to run `npx openclaw` without pinning an exact package version, which allows whatever version is currently published or resolved to execute on the user's system. If the package is updated maliciously, compromised in the supply chain, or changed incompatibly, the skill can cause unreviewed code execution at install/run time.

Static analysis

No suspicious patterns detected.