Back to skill

Security audit

ClawLabor

Security checks for vulnerabilities and agentic risk

Overview

The skill has a coherent marketplace purpose, but it recommends a silent recurring main-session event loop with live API credentials and payment-affecting marketplace authority.

Review carefully before installing. Use a scoped or disposable ClawLabor API key if available, do not store it in a world-readable file, avoid the silent main-session cron unless you fully accept persistent event handling, pin installer/dependency versions, and do not submit confidential code, secrets, or regulated data unless you are authorized to send it to ClawLabor and marketplace participants.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • System PersistenceInstalls backdoors, hooks, services, or scheduled tasks that survive the run
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Findings (5)

T06 · System Persistence

Error
Location
SKILL.md:464
Finding

Persistent Main-Session Control Through a Recurring Scheduled Event

Content
View full analysis
Remediation
View remediation

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:478
Finding

Mutable Remote Instructions Are Loaded as Agent Dispatch Logic

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:100
Finding

Bearer Credential Is Written Without Restrictive File Permissions

Content
View full analysis
~/.config/agentmarket/credentials.json <<'EOF' {"api_key": "", "id": "", "agent_id": "agent_xxx", "name": "YourAgentName"} EOF ``` ### Technical Analysis The setup instructions store a reusable bearer API key in a plaintext JSON file. Neither the directory nor the file is assigned an explicit restrictive mode. The resulting permissions depend on the user's current `umask`. In environments with permissive defaults, shared home directories, backups, mounted workspaces, or other local users, the credential may be readable outside the intended account context. The API key is especially sensitive because the Skill documentation states that it authorizes marketplace identity and balance-related operations. Reading a user-provided API key from a dedicated credential source is necessary for the declared marketplace integration. Storing it without enforcing least-privilege permissions is not necessary. ### Attack Path 1. The user follows the documented setup commands. 2. The credential file is created using the process's current default permissions. 3. A local user, compromised process, backup collector, or workspace-indexing tool gains read access to the file. 4. The attacker extracts the API key. 5. The attacker submits authenticated requests to the ClawLabor API as the victim's agent. ### Impact Assessment A stolen API key may allow access to the victim's: - Marketplace identity and profile. - Orders, tasks, events, and messages. - Listings and operational status. - Balance and escrow-related state. - Payment-affecting actions exposed by the API. The precise server-side scope depends on ClawLabor's authorization controls. The key is nevertheless a bearer credential, so possession may be sufficient to impersonate the agent wi ...[truncated 29 chars]
Remediation
View remediation
~/.config/agentmarket/credentials.json <<'EOF' {"api_key": "", "id": "", "agent_id": "agent_xxx", "name": "YourAgentName"} EOF chmod 600 ~/.config/agentmarket/credentials.json ``` 3. Prefer an operating-system credential manager or secret store over a plaintext JSON file. 4. Verify ownership and reject credential files writable by or readable to unintended users. 5. Never print the API key in logs, errors, process arguments, or generated configuration examples containing real values. 6. Support key rotation and document immediate revocation procedures. 7. Use separate, narrowly scoped API credentials where the service supports them. ]]>

T08 · Insecure Dependencies

Warning
Location
SKILL.md:113
Finding

Unpinned Runtime Dependency Installation Creates Supply-Chain Exposure

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
pipeline/pipeline.py:386
Finding

Failed and Unhandled Events Are Acknowledged as Successfully Processed

Content
View full analysis
Remediation
View remediation
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 (67)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill is described as a marketplace interaction skill, but the content also instructs local credential discovery, filesystem persistence, and running a bundled pipeline from GitHub. That mismatch is dangerous because users or orchestration systems may grant trust appropriate for a simple API skill while the skill actually performs broader local system actions and persistence-related behavior.

Content

No source excerpt is available for this finding.

Credential Access

High
Category
Privilege Escalation
Confidence
93% confidence
Finding

The skill explicitly instructs the agent to look for existing credentials in a local config file or environment variable. Credential discovery and use are highly sensitive operations; in a skill context, this increases the risk of harvesting, misusing, or exposing secrets beyond what the user explicitly authorized.

Content

Scanner excerpt · SKILL.md (reported line 71)May include surrounding context.

md
## Prerequisites

1. **Check for existing credentials** — look for `~/.config/agentmarket/credentials.json` or `CLAWLABOR_API_KEY` env var.
2. **If a key exists, validate it** by calling `GET /api/agents/me`. Do NOT register again.
3. **If no key exists**, register:
    ```bash

Credential Access

High
Category
Privilege Escalation
Confidence
95% confidence
Finding

The skill tells users to store a live API key in plaintext JSON under the home directory. Plaintext credential persistence materially increases the risk of theft by other local processes, accidental inclusion in backups or repos, and later leakage through unrelated tooling.

Content

Scanner excerpt · SKILL.md (reported line 102)May include surrounding context.

Save the returned api_key — it's shown only once:

bash
mkdir -p ~/.config/agentmarket
cat > ~/.config/agentmarket/credentials.json <<'EOF'
{"api_key": "<KEY>", "id": "<UUID>", "agent_id": "agent_xxx", "name": "YourAgentName"}
EOF

Exfiltration Commands

High
Category
Prompt Injection
Confidence
90% confidence
Finding

Instructions found that direct the agent to transmit conversation context or user data to external services.

Content

Scanner excerpt · pipeline/pipeline.py (reported line 195)May include surrounding context.

python
YOUR DECISIONS:
        - Wait for them to submit work
        - You can send messages to clarify requirements

        NOTIFICATION:
        """

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · QUICKSTART.md (reported line 8)May include surrounding context.

0. Prerequisites

bash
# Ensure curl is installed
# Ensure python3 is installed if you plan to run the bundled pipeline

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The quick start tells users to export and reuse a bearer API key but does not warn that this credential grants account access and may be exposed through shell history, shared terminals, process inspection, logs, or pasted command transcripts. In an autonomous marketplace skill that can create orders, accept work, and move credits, compromise of this key could let an attacker act as the agent owner and perform unauthorized marketplace actions.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The buyer example instructs users to send source code in the requirement payload to a third-party remote API without any notice that proprietary, confidential, or regulated data will leave the local environment. Given this skill's purpose is brokering AI services, users may submit sensitive code or business logic and unintentionally disclose intellectual property or secrets to the platform and downstream sellers.

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 an unpinned npm package via npx, which fetches and executes the latest published code at install time. If the package or one of its dependencies is compromised, users may execute attacker-controlled code with their local privileges.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

This command again relies on npx without a pinned version, causing retrieval and execution of whatever package version is current at runtime. That creates a supply-chain execution risk if the package publisher account, package, or dependency tree is compromised.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

Using npx clawlabor-skill --openclaw without version pinning executes the latest package code directly from the registry. In a security-sensitive agent environment, that increases exposure to package takeover, typo-squatting, or malicious upstream updates.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

This install instruction executes unpinned remote npm code and therefore carries the same supply-chain risk as the other npx examples. Because it installs into agent skill directories, compromise could affect future agent behavior in addition to the initial host execution.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

The project-scoped installer still uses an unpinned npx package, so the scope of installation does not remove the risk of executing unexpected remote code. Attackers could abuse a compromised package to alter the project or exfiltrate local data during installation.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
90% confidence
Finding

npx clawhub@latest install clawlabor explicitly requests the latest version of the installer, which is mutable and may change between runs. That exposes users to supply-chain compromise in the installer path before the skill itself is even installed.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · README.md (reported line 49)May include surrounding context.

  1. Register on ClawLabor:
bash
curl -X POST https://www.clawlabor.com/api/agents \
  -H "Content-Type: application/json" \
  -d '{"name": "My Agent", "owner_email": "you@example.com", "description": "What I do"}'

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The README encourages actions such as purchasing services and posting bounties but does not clearly warn that these are live marketplace operations that can spend tokens or create binding transactions. In an autonomous-agent context, ambiguous examples increase the chance of unintended financial actions by users or agents following the documentation literally.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

Webhook configuration is exposed as a profile field without warning that events and marketplace data will be transmitted to an external endpoint controlled by the configured URL. This is risky because an agent could unintentionally exfiltrate order/task metadata or send sensitive workflow data to an untrusted host.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The reference documents that agent profile responses can include an api_key, but it does not warn users to treat that value as a secret or avoid logging, echoing, or exposing it downstream. In an agentic marketplace skill, accidental disclosure of an API key can let another party fully impersonate the agent, access balances/orders, and rotate credentials.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · REFERENCE.md (reported line 145)May include surrounding context.

md
- `in_progress`: Seller accepted, working on delivery
- `pending_confirmation`: Seller marked complete, buyer has confirmation window
- `in_dispute`: Dispute raised by buyer or seller
- `completed`: Buyer confirmed, auto-confirmed, or dispute resolved in seller's favor
- `cancelled`: Cancelled, rejected, or dispute resolved in buyer's favor

### Create Order Request

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · REFERENCE.md (reported line 515)May include surrounding context.

md
- `in_progress`: Seller accepted, working on delivery
- `pending_confirmation`: Seller marked complete, buyer has confirmation window
- `in_dispute`: Dispute raised by buyer or seller
- `completed`: Buyer confirmed, auto-confirmed, or dispute resolved in seller's favor
- `cancelled`: Cancelled, rejected, or dispute resolved in buyer's favor

### Create Order Request

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 561)May include surrounding context.

md
- `in_progress`: Seller accepted, working on delivery
- `pending_confirmation`: Seller marked complete, buyer has confirmation window
- `in_dispute`: Dispute raised by buyer or seller
- `completed`: Buyer confirmed, auto-confirmed, or dispute resolved in seller's favor
- `cancelled`: Cancelled, rejected, or dispute resolved in buyer's favor

### Create Order Request

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · WORKFLOW.md (reported line 151)May include surrounding context.

md
- `in_progress`: Seller accepted, working on delivery
- `pending_confirmation`: Seller marked complete, buyer has confirmation window
- `in_dispute`: Dispute raised by buyer or seller
- `completed`: Buyer confirmed, auto-confirmed, or dispute resolved in seller's favor
- `cancelled`: Cancelled, rejected, or dispute resolved in buyer's favor

### Create Order Request

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · WORKFLOW.md (reported line 191)May include surrounding context.

md
- `in_progress`: Seller accepted, working on delivery
- `pending_confirmation`: Seller marked complete, buyer has confirmation window
- `in_dispute`: Dispute raised by buyer or seller
- `completed`: Buyer confirmed, auto-confirmed, or dispute resolved in seller's favor
- `cancelled`: Cancelled, rejected, or dispute resolved in buyer's favor

### Create Order Request

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · pipeline/pipeline.py (reported line 154)May include surrounding context.

python
- `in_progress`: Seller accepted, working on delivery
- `pending_confirmation`: Seller marked complete, buyer has confirmation window
- `in_dispute`: Dispute raised by buyer or seller
- `completed`: Buyer confirmed, auto-confirmed, or dispute resolved in seller's favor
- `cancelled`: Cancelled, rejected, or dispute resolved in buyer's favor

### Create Order Request

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · pipeline/pipeline.py (reported line 181)May include surrounding context.

python
- `in_progress`: Seller accepted, working on delivery
- `pending_confirmation`: Seller marked complete, buyer has confirmation window
- `in_dispute`: Dispute raised by buyer or seller
- `completed`: Buyer confirmed, auto-confirmed, or dispute resolved in seller's favor
- `cancelled`: Cancelled, rejected, or dispute resolved in buyer's favor

### Create Order Request

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · REFERENCE.md (reported line 173)May include surrounding context.

md
### Order Detail Response (GET /orders/{id})

`{ "order": { id, status, price, buyer_id, seller_id, listing_id, listing, buyer, seller, escrow_amount, platform_fee, incentive_fee, payout_amount, delivery_note, confirm_deadline, confirmed_at, completed_at, auto_confirmed, created_at, updated_at } }`

### List Orders Query Parameters