Back to skill

Security audit

WoClaw

Security checks for vulnerabilities and agentic risk

Overview

WoClaw is a coherent shared-memory and messaging skill, but its docs expose users to review-worthy risks from plaintext hub connections, shared global memory, and unpinned external installs.

Review before installing. Use a pinned, verified package and container digest; prefer wss:// and https:// with valid TLS for any non-local hub; do not place secrets, credentials, private prompts, or sensitive personal data in shared memory or topic messages; and limit hub tokens, topics, and memory access to trusted agents only.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (2)

T08 · Insecure Dependencies

Error
Location
README.md:47
Finding

Unpinned Third-Party Packages and Mutable Container Image

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:24-29; README.md:19-23; README.md:47-60
Vulnerability Type: Supply-chain exposure through unpinned external artifacts
Risk Level: High

Vulnerable Code

SKILL.md:24-29:

bash
### 1. Install the Skill

```bash
npx clawhub install woclaw
text

`README.md:19-23`:

```bash
## Quick Start

```bash
npx clawhub install woclaw
text

`README.md:47-60`:

```bash
Deploy your own Hub:

```bash
docker run -d \
  --name woclaw-hub \
  -p 8082:8082 -p 8083:8083 \
  -e AUTH_TOKEN=your-token \
  xingp14/woclaw-hub:latest

Or via Node.js:

bash
npm install -g xingp14-woclaw
WOCLAW_AUTH_TOKEN=your-token woclaw hub
text

### Technical Analysis

The installation instructions retrieve and execute third-party artifacts without pinning immutable versions or integrity hashes:

- `npx clawhub install woclaw` does not identify a reviewed, exact Skill version.
- `npm install -g xingp14-woclaw` installs the registry-selected package version globally.
- `xingp14/woclaw-hub:latest` is a mutable container tag whose content can change after this audit.

Consequently, the code users execute can differ from the artifact that was reviewed. A compromised publisher account, package registry, container registry, or later malicious release could replace the effective implementation while the documented command remains unchanged. Global package installation can also expose a malicious package lifecycle script or executable to the invoking user's privileges.

The audited project contains only documentation and does not include the downloaded package or container implementation. Therefore, their behavior and integrity cannot be established from this project.

### Attack Path

1. An attacker compromises the relevant package publisher, registry account, image repository, or release process.
2. The attacker publishes a malicious p
...[truncated 1220 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin the Skill and npm package to exact, reviewed versions rather than relying on registry-selected releases.
  2. Pin the container image by immutable digest, for example:
    bash
    docker run ... xingp14/woclaw-hub@sha256:<verified-digest>
    
  3. Publish and verify cryptographic checksums, signatures, provenance attestations, or Sigstore signatures for every release.
  4. Avoid global npm installation where possible. Use a project-local, locked dependency with a committed lockfile.
  5. Review package lifecycle scripts before installation and use installation controls that disable scripts when they are unnecessary.
  6. Run the Hub as a non-root user with a read-only filesystem, minimal Linux capabilities, restricted network access, and no unnecessary host mounts.
  7. Document a tested upgrade process so version and digest changes require explicit security review.
  8. Add automated dependency, container, and provenance scanning to the release pipeline.

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:43
Finding

Authentication and Agent Data Are Documented Over Plaintext Network Protocols

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:43-48; SKILL.md:57-64; SKILL.md:139-145; SKILL.md:189-200; README.md:12-15; README.md:35-41
Vulnerability Type: Plaintext transmission of credentials and sensitive agent data
Risk Level: High

Vulnerable Code

SKILL.md:43-48:

bash
### 3. Configure environment variables

```bash
export WOCLAW_HUB_URL=ws://your-hub-host:8082
export WOCLAW_AGENT_ID=your-agent-name
export WOCLAW_TOKEN=your-token
text

`SKILL.md:57-64`:

```json
{
  "channels": {
    "woclaw": {
      "enabled": true,
      "hubUrl": "ws://your-hub-host:8082",
      "agentId": "your-agent-name",
      "token": "your-token",
      "autoJoin": ["general", "openclaw-help"]

SKILL.md:139-145:

text
| Option | Environment Variable | Default | Description |
|--------|---------------------|---------|-------------|
| `hubUrl` | `WOCLAW_HUB_URL` | `ws://localhost:8082` | WoClaw Hub WebSocket URL |
| `agentId` | `WOCLAW_AGENT_ID` | `openclaw` | Your agent's unique ID |
| `token` | `WOCLAW_TOKEN` | (required) | Authentication token |
| `autoJoin` | `WOCLAW_AUTO_JOIN` | `[]` | Topics to join on startup |

SKILL.md:189-200:

text
## External Endpoints

| Endpoint | Type | Description |
|----------|------|-------------|
| `WOCLAW_HUB_URL` (WS) | WebSocket | WoClaw Hub WebSocket for real-time messaging |
| `WOCLAW_HUB_URL:8083` | REST | WoClaw Hub REST API for memory and topic operations |

## Security & Privacy

- **Data transmitted**: Agent ID, topic messages, and shared memory content are sent to the WoClaw Hub
- **Authentication**: Token-based auth (`WOCLAW_TOKEN`) required for Hub connection
- **Network access**: Requires outbound WebSocket (port 8082) and HTTP (port 8083) access to Hub host
- **No external data collection**: This skill does not send data to any third-party services beyond your configured WoClaw Hub
...[truncated 3140 chars]
Remediation
View remediation

Remediation Suggestions

  1. Require wss:// for WebSocket connections and https:// for REST endpoints in production.
  2. Reject plaintext endpoints by default. If plaintext is retained for local development, restrict it explicitly to loopback addresses and display a prominent warning.
  3. Configure valid server certificates and enforce hostname and certificate-chain validation.
  4. Do not provide an insecure ws:// production example or default. Change the examples and configuration table to secure URLs.
  5. Use short-lived, scoped tokens with separate permissions for topics, memory operations, and administration.
  6. Rotate any token that may previously have been transmitted over plaintext transport.
  7. Store tokens in an operating-system secret store or dedicated secrets manager rather than plaintext configuration where supported.
  8. Add mutual TLS or another strong client-authentication layer for sensitive deployments.
  9. Apply per-topic and per-key authorization instead of relying solely on possession of a global token.
  10. Add audit logs, replay protections, rate limits, and rapid token-revocation capabilities.
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 (5)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The README promotes shared memory and multi-agent topic messaging but does not clearly warn that data written there is globally accessible to other connected agents and may influence their behavior. This is especially risky in an AI-agent context because operators may unknowingly place secrets, prompts, or decision-relevant state into a shared channel where other agents can read, misuse, or poison it.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

The README instructs users to run npx clawhub install woclaw without pinning a specific package version. This creates a supply-chain risk because users may fetch a newer or compromised package than the author tested, and npx executes retrieved code directly. In a skill that facilitates agent messaging and shared memory, compromise could affect agent coordination and expose sensitive context.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The Docker guidance uses the mutable latest tag (xingp14/woclaw-hub:latest), which means users may deploy an unexpected or tampered image over time. Mutable container references are a well-known supply-chain hazard, and here the hub handles shared memory, agent messaging, and authentication tokens, so a malicious image could intercept or manipulate multi-agent traffic.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The install instruction uses npx clawhub without a pinned version, which means users may execute whatever package version is current at install time. Because npx fetches and runs code, a compromised upstream release, typo-squatted package, or breaking update could result in unexpected code execution in the user's environment.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill documents /woclaw memory delete <key> even though shared memory is described as global and accessible by all connected agents, but it does not warn that deletion may affect other agents or be irreversible. In a multi-agent shared state system, this omission increases the chance of accidental destructive actions, causing coordination failures or loss of shared information.

Content

No source excerpt is available for this finding.