Back to skill

Security audit

Clawmart Upload

Security checks for vulnerabilities and agentic risk

Overview

This skill is purpose-built to upload OpenClaw packs, but it collects and uploads broad local agent configuration and stores an API token in plaintext with limited safeguards.

Review every file before uploading, especially memory, identity, user, tools, bootstrap, and local skill files. Do not use the broad include-all path unless you are comfortable sending that content to ClawMart, and avoid pasting or storing a token unless you can protect or revoke it.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • 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 (4)

T05 · Unauthorized Access and Privilege Escalation

Error
Location
SKILL.md:49
Finding

Broad Collection and External Upload of Sensitive OpenClaw State

Content
View full analysis
", "content": "" } ``` Use just the filename (no path prefix) — e.g., `"SOUL.md"`, `"AGENTS.md"`, `"memory_projects.json"`. ``` ```markdown ## Step 6: Upload to ClawMart Send the upload request: POST {base_url}/api/packs Authorization: Bearer {token} Content-Type: application/json { "title": "", "description": "", "version": "", "files": [ ...files array from Step 5... ] } ``` ### Technical Analysis The skill directs the agent to read complete OpenClaw files containing identity, memory, user information, tool configuration, bootstrap instructions, and behavioral state. It then places their full contents into a request sent to `https://clawmart-gray.vercel.app/api/packs`. Although ...[truncated 1589 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:89
Finding

Local Skill Files Are Uploaded Without Sensitive-Information Scanning

Content
View full analysis
Sensitive information detected in {filename} at line {line}: `{masked_value}`. It is recommended to remove it before uploading. Continue anyway? (y/n) Only proceed if user says yes. ``` ```markdown 2. For each **local skill** (user-authored, no `_meta.json`) confirmed in Step 2, add: ```json { "name": "skills/", "content": "" } ``` Preserve the `skills/` prefix so the server can classify them correctly. ``` ### Technical Analysis The workflow explicitly limits sensitive-information scanning to “non-SKILLS files,” while separately directing the agent to upload the full contents of locally authored `SKILL.md` files. Local skills can contain API tokens, internal URLs, credentials in examples, private instructions, or environment-specific information. Because these files are excluded from the only prescribed security scan, their contents can enter the upload payload without warning. This creates a direct validation gap between the collection and transmission stages. ### Attack Path 1. A local skill contains a token, password, private endpoint, or other sensitive value. 2. The skill slug is absent ...[truncated 869 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:107
Finding

Sensitive-Information Detection Uses Narrow and Bypassable Patterns

Content
View full analysis
Sensitive information detected in {filename} at line {line}: `{masked_value}`. It is recommended to remove it before uploading. Continue anyway? (y/n) Only proceed if user says yes. ``` ### Technical Analysis The prescribed patterns recognize only a limited set of token prefixes and simple one-line credential assignments. They do not cover many common secret formats, including: - PEM or SSH private keys. - JSON Web Tokens. - Cloud-provider access keys and service-account documents. - Database connection strings. - Session cookies. - Credentials containing spaces, quotes, or multiline values. - Vendor tokens with unlisted prefixes. - High-entropy secrets without a recognizable prefix. The control is also fail-open: even when a sensitive value is detected, the user can authorize uploading the original unredacted value. Users may not understand the exposure implications or may approve the prompt reflexively. ### Attack Path 1. A selected configuration file contains a credential in an unsupported format. 2. The narrow regular expressions do not match the credential. 3. No warning is displayed. 4. The complete file is added to the upload payload. 5. The payload is sent to the external marketplace with the credential intact. Alternatively, a supported secret is detected, but the user accepts the continuation prompt ...[truncated 473 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:25
Finding

Reusable API Token Is Persisted Without Filesystem Protection Requirements

Content
View full analysis
You need a ClawMart API Token to upload. Please visit https://clawmart-gray.vercel.app/dashboard/tokens to generate one, then paste it here. Once the user provides a token (format: `cm_` followed by hex characters), save it: ```json { "token": "", "base_url": "https://clawmart-gray.vercel.app" } ``` Write this to `~/.openclaw/clawmart-config.json`. ``` ### Technical Analysis The skill requires a reusable bearer token to be stored in a plaintext JSON file but does not specify restrictive permissions, secure file creation, ownership validation, atomic replacement, or use of an operating-system credential store. The actual exposure depends on the host's default permissions and directory configuration. If the file is created with permissive access, another local user or unrelated process may read and reuse the token. Backups, synchronization tools, or diagnostic collection may also copy the plaintext credential. ### Attack Path 1. The user supplies a valid ClawMart API token. 2. The agent writes the token to `~/.openclaw/clawmart-config.json`. 3. The file is created using inherited or permissive filesystem permissions. 4. Another local principal, process, backup service, or synchronization utility reads the file. 5. The exposed bearer token is used to authenticate to ClawMart as the victim. ### Impact Assessment An attacker who obtains the token gains the ClawMart permissions associated with that credential. Depending on server-side authorization, this may permit pack submission, access to account resources, modification of marketplace content, or other authenticated operations. Th ...[truncated 172 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 72)May include surrounding context.

md
- **Slug is NOT in `lock.json`** → user-authored locally (never installed from clawhub). Include the full `SKILL.md` content in the zip under `skills/`.

Ssd 3

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill instructs the agent to include full contents of non-skill configuration files and local skills directly in the upload payload. This creates a direct exfiltration path for sensitive prompts, memory files, user instructions, internal tooling definitions, and other private data; the included sensitive-pattern scan is narrow and may miss many secrets or confidential business information.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The trigger "upload my pack" is generic and lacks product or context qualifiers, unlike the other ClawMart-specific triggers. This increases the chance of unintended invocation when a user is talking about uploading some other kind of pack.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill begins a workflow that reads and uploads the user's OpenClaw configuration to a remote service, but it does not present an upfront privacy warning before scanning files and collecting content. Although it later asks for confirmation on included files, users are not clearly warned at the outset that configuration text and local skill contents may be transmitted off-device.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
84% confidence
Finding

The skill stores the ClawMart API token in a persistent local configuration file for future reuse. Persisting bearer tokens increases exposure if the local environment is compromised, backups are leaked, or file permissions are weak, and the skill provides no safeguards such as restrictive permissions, OS keychain usage, or token scoping guidance.

Content

Scanner excerpt · SKILL.md (reported line 40)May include surrounding context.

}

text

Write this to `~/.openclaw/clawmart-config.json`.

---