Back to skill

Security audit

AIOT Network

Security checks for vulnerabilities and agentic risk

Overview

This skill is a visible router for AIOT financial and identity skills, but it asks users to install several sensitive external skills without version or publisher safeguards.

Review this before installing. The skill itself is a disclosed router and installer, not an observed data stealer or destructive payload, but it can install and delegate to several external skills for identity, payment, card, crypto, and account operations. Only install it if you trust the publisher and the exact downstream skills, and require explicit confirmation before document uploads, payments, withdrawals, card changes, staking, or other account-impacting actions.

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

Error
Location
scripts/install.sh:7
Finding
Unpinned External Skills Create a Supply-Chain Compromise Risk## Vulnerability Details **File Location**: `scripts/install.sh:7-24` **Vulnerability Type**: Unpinned and unverified third-party skill installation **Risk Level**: High The installer downloads six external ClawHub skills using only mutable package slugs. It does not pin versions or cryptographic digests and does not verify package signatures, publisher identities, or downloaded contents. ```bash SKILLS=( "aiotnetwork-account-auth" "aiotnetwork-kyc-identity" "aiotnetwork-card-management" "aiotnetwork-payments-banking" "aiotnetwork-crypto-wallet" "aiotnetwork-blockchain-did" ) echo "Installing ${#SKILLS[@]} AIOT Network skills..." echo "" FAILED=0 SUCCEEDED=0 for slug in "${SKILLS[@]}"; do echo "→ Installing ${slug}..." if clawhub install "${slug}"; then ``` The same unpinned installation approach is documented in `SKILL.md:62-84`, including both the bulk installation script and individual `clawhub install` commands. ### Technical Analysis Package slugs generally identify mutable registry entries rather than immutable, audited artifacts. Consequently, the code installed by this script can change after this project has been reviewed. No control in the supplied project ensures that future installations retrieve the same dependency content that was originally intended. The external skills are responsible for authentication, identity verification, card management, banking, payments, and cryptocurrency operations. Their source code is not present in this project and therefore was outside the available audit scope. This finding does not establish that any current dependency is malicious; it establishes that the installation process provides no protection against a dependency becoming malicious or being replaced. ### Attack Path 1. An attacker compromises a listed ClawHub package, its publisher account, or the relevant registry distribution path. 2. The attacker publishes modified content ...[truncated 1530 chars]
Remediation
## Remediation Suggestions 1. Pin every external skill to an explicitly reviewed, immutable version. 2. Where supported, pin and verify a cryptographic digest for each downloaded artifact. 3. Require valid package signatures and verify that each signature belongs to an allowlisted publisher identity. 4. Fail closed if a version, digest, signature, or publisher check cannot be completed. 5. Maintain a lock file or manifest recording each slug, exact version, digest, publisher, and review date. 6. Audit and test the exact dependency artifacts before allowing them to process identity, payment, card, or cryptocurrency data. 7. Use a trusted internal mirror or vendor reviewed dependency contents when immutable ClawHub installation is unavailable. 8. Run installed skills with least privilege, isolated credentials, restricted filesystem and network access, and explicit approval for sensitive transactions. 9. Add automated dependency monitoring and require security review before changing any pinned version or digest.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding
The file is presented as a routing/index skill, but it also instructs users to run shell commands and install external skills, while providing no actual routing logic beyond prose. That mismatch is dangerous because agents or operators may trust the declared purpose and overlook that the content can trigger environment changes or pull additional code from external packages, expanding the attack surface through implicit supply-chain and execution steps.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
This skill routes users into high-risk domains including payments, identity verification, card issuance, and cryptocurrency withdrawal, yet it provides no warning about handling sensitive personal data, financial consequences, or irreversible blockchain actions. In this context, the omission increases the chance of unsafe delegation, uninformed user actions, and accidental disclosure or execution in sensitive workflows.

Static analysis

No suspicious patterns detected.