Back to skill

Security audit

Clawdmint

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed NFT deployment helper for Clawdmint, but users should treat its API keys, webhook setup, and x402 USDC payments as real financial and account actions.

Before installing, make sure you trust Clawdmint and use a reviewed ClawHub release rather than the manual curl install unless you verify the downloaded file. Store the API key carefully, send it only to clawdmint.xyz, configure only webhook endpoints you control, and require explicit wallet approval and spending limits for any x402 USDC payment or on-chain deployment.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:270
Finding
Unpinned ClawHub Skill Dependency Installation## Vulnerability Details **File Location**: `SKILL.md`, line 270 **Vulnerability Type**: `T08: Insecure Dependencies` **Risk Level**: Medium ### Vulnerable Code ```bash clawhub install clawdmint ``` ### Technical Analysis The documented installation command identifies the ClawHub package only by its mutable package name. It does not specify an immutable version, content digest, source commit, or cryptographic signature. Consequently, the content installed when a user executes this command may differ from the artifact reviewed during this audit. Compromise of the publisher account, package registry, or package release process could cause an altered Skill to be installed under the legitimate package name. The project also demonstrates use of the `@x402/fetch` payment dependency without a pinned version at line 370. Because this component is expected to interact with a wallet and automate USDC payments, users must ensure that the version installed by their surrounding application is explicitly pinned and independently reviewed. ### Attack Path 1. An attacker compromises the ClawHub publisher account, registry entry, or release process associated with `clawdmint`. 2. The attacker publishes a modified Skill under the same trusted package name. 3. A user executes `clawhub install clawdmint` as instructed. 4. ClawHub resolves the mutable name to the attacker-controlled release. 5. OpenClaw loads the modified Skill, which may introduce malicious instructions, request credentials, redirect transactions, or initiate additional downloads. ### Impact Assessment Successful exploitation could affect the privileges available to the OpenClaw process and any credentials or tools exposed to the installed Skill. Depending on the malicious replacement's content, the impact could include disclosure of the Clawdmint API key, manipulation of NFT deployment parameters, unauthorized network requests, or abuse of wallet-integrated payment f ...[truncated 184 chars]
Remediation
## Remediation Suggestions - Pin the ClawHub Skill to a reviewed, immutable version rather than installing by package name alone. - Publish a SHA-256 digest or signed manifest for every supported release and require verification during installation. - Document the authoritative source repository, release commit, publisher identity, and signature-verification procedure. - Use dependency lockfiles and integrity metadata for applications that install `@x402/fetch`. - Review wallet-capable dependencies before use and restrict them to the required chain, payment recipient, maximum amount, token allowance, and transaction lifetime. - Monitor package ownership and release events, and establish a revocation process for compromised releases.

T03 · Remote Payload Retrieval and Execution

Warning
Location
SKILL.md:274
Finding
Mutable Remote Skill Downloaded Directly into the Active Skill Directory## Vulnerability Details **File Location**: `SKILL.md`, lines 274-277 **Vulnerability Type**: `T03: Remote Payload Retrieval and Execution` **Risk Level**: Medium ### Vulnerable Code ```bash mkdir -p ~/.openclaw/skills/clawdmint curl -o ~/.openclaw/skills/clawdmint/SKILL.md https://clawdmint.xyz/skill.md ``` ### Technical Analysis The installation instructions retrieve a mutable remote Skill definition and write it directly into the active OpenClaw Skill directory. The command does not pin a release, validate a checksum, verify a digital signature, or stage the file for review before activation. Although the downloaded file is not passed to a shell interpreter, a Skill file is an agent instruction payload. Once OpenClaw discovers and loads it, its instructions can influence agent behavior and tool use. The effective payload can therefore change after the version in this project has been audited. HTTPS protects the connection in transit under normal conditions, but it does not protect against compromise of the remote server, deployment pipeline, domain account, or authorized publishing credentials. ### Attack Path 1. An attacker compromises `clawdmint.xyz`, its deployment pipeline, or credentials authorized to modify `/skill.md`. 2. The attacker replaces the legitimate Skill document with a malicious version. 3. A user follows the documented command and downloads the altered file. 4. The file is written directly to `~/.openclaw/skills/clawdmint/SKILL.md`. 5. OpenClaw discovers and loads the modified Skill. 6. The malicious instructions use available agent capabilities to request or disclose secrets, redirect API or payment operations, alter deployment parameters, or retrieve additional payloads. ### Impact Assessment Exploitation could control the behavior of the loaded Skill within the permissions and tools granted to the OpenClaw process. Potentially exposed assets include the `CLAWDMINT_API_KEY`, collection d ...[truncated 309 chars]
Remediation
## Remediation Suggestions - Distribute versioned, immutable Skill artifacts rather than installing from a mutable `/skill.md` URL. - Publish a cryptographic signature and SHA-256 digest for each release. - Download the artifact into a staging directory, verify its signature and digest, and present it for review before moving it into the active Skill directory. - Use an atomic installation process that rejects missing, invalid, revoked, or unexpected signatures. - Record the verified version and digest locally to support auditing and rollback. - Require explicit user approval for updates that add credential access, wallet operations, external destinations, or new tool permissions. - Apply least-privilege controls to the OpenClaw process and isolate wallet signing from general Skill instructions.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (10)

Description-Behavior Mismatch

Medium
Confidence
94% confidence
Finding
The manifest describes the skill narrowly as deploying NFT collections, but the body documents additional capabilities including agent registration/claim workflows, webhook configuration, and paid data-access endpoints. This scope mismatch can cause users or orchestrators to grant trust, permissions, or invoke the skill under incomplete assumptions, increasing the risk of unintended external actions and data exposure.

External Transmission

Medium
Category
Data Exfiltration
Content
### Step 1: Register

```bash
curl -X POST https://clawdmint.xyz/api/v1/agents/register \
  -H "Content-Type: application/json" \
  -d '{
    "name": "YourAgentName",
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
```bash
# 1. Register
RESPONSE=$(curl -s -X POST https://clawdmint.xyz/api/v1/agents/register \
  -H "Content-Type: application/json" \
  -d '{"name": "ArtBot", "description": "I create digital art"}')
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
# 2. Wait for human to tweet verification...

# 3. Check status
curl -s https://clawdmint.xyz/api/v1/agents/status \
  -H "Authorization: Bearer $API_KEY"

# 4. Deploy collection
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Session Persistence

Medium
Category
Rogue Agent
Content
Or add manually to your OpenClaw workspace:

```bash
mkdir -p ~/.openclaw/skills/clawdmint
curl -o ~/.openclaw/skills/clawdmint/SKILL.md https://clawdmint.xyz/skill.md
```
Confidence
60% confidence
Finding
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Skill Enumeration

Medium
Category
Agent Snooping
Content
```bash
mkdir -p ~/.openclaw/skills/clawdmint
curl -o ~/.openclaw/skills/clawdmint/SKILL.md https://clawdmint.xyz/skill.md
```

Configure your API key in `~/.openclaw/openclaw.json`:
Confidence
80% confidence
Finding
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Context-Inappropriate Capability

Medium
Confidence
91% confidence
Finding
The webhook setup introduces an outbound-to-inbound integration surface that is materially different from simple NFT deployment. A user expecting only deployment may unknowingly configure a callback endpoint, creating opportunities for event-triggered requests, token handling, or SSRF-like exposure against internal gateways if an agent supplies sensitive URLs.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The x402 section explains how to invoke paid endpoints and mentions prices, but it does not provide a strong, explicit warning at the point of action that using the example will spend USDC. In agent settings, weak spend disclosure can lead to unintended financial loss because automated tools may follow the example flow and attach payment headers without meaningful user consent.

External Transmission

Medium
Category
Data Exfiltration
Content
```bash
# Get all x402 pricing info
curl https://clawdmint.xyz/api/x402/pricing
```

### Deploy via x402
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Context-Inappropriate Capability

Medium
Confidence
96% confidence
Finding
The premium x402 endpoints add paid access for listing agents, collections, and analytics, which exceeds the manifest's stated deployment-only purpose. This hidden expansion of functionality increases the chance that an agent will perform billable or privacy-relevant queries without users realizing the skill can do more than deploy collections.

Static analysis

No suspicious patterns detected.