Back to skill

Security audit

Mem0 Tech Tree Memory

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent local memory tool, but it ships real-looking personal memory records in plaintext and has under-scoped persistence and deletion behavior that users should review before installing.

Review tree.json before installing. Remove the bundled memory database or replace it with empty or synthetic data, keep runtime memories outside the published package, avoid storing secrets or private account details, add confirmation or backup for clear, and remove or pin the unused mem0ai dependency.

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)

T09 · Insecure Skill Coding Practices

Warning
Location
tree.json:817
Finding

Plaintext Distribution of Personal and Environment Information

Content
View full analysis

Vulnerability Details

File Location: tree.json:817-839, tree.json:992-1014, tree.json:1223-1246
Vulnerability Type: Plaintext sensitive information exposure
Risk Level: Medium

The distributed memory database contains personal, social-media, employment, location, and device-profile information in plaintext.

json
"n0033": {
  "id": "n0033",
  "content": "抖音阿拉口木(966403340)2700粉,2月停更",
  "keywords": [
    "抖音",
    "阿拉",
    "口木",
    "966403340",
    "2700"
  ],
  "branches": [
    "内容创作",
    "生活"
  ],
  "tier": 2,
  "state": "unlocked",
  "importance": "high",
  "types": [
    "semantic"
  ],
  "access_count": 3
}
json
"n0040": {
  "id": "n0040",
  "content": "大先生:47岁扬州销售管理,高毛品/订单/促销",
  "keywords": [
    "先生",
    "47",
    "扬州",
    "销售",
    "管理",
    "高毛品",
    "订单",
    "促销"
  ],
  "branches": [
    "生活",
    "工作"
  ],
  "tier": 2,
  "state": "unlocked",
  "importance": "high"
}
json
"n0050": {
  "id": "n0050",
  "content": "本机为Intel i9+16GB+核显1.5GB,不适合本地SD/ComfyUI,继续用即梦AI云端方案",
  "keywords": [
    "本机为Intel",
    "GB",
    "核显",
    "GB",
    "不适合本地SD",
    "ComfyUI",
    "继续用即梦AI云端方案"
  ],
  "branches": [
    "技术",
    "内容创作"
  ],
  "tier": 2,
  "state": "unlocked",
  "importance": "high"
}

Technical Analysis

tree.json is included directly in the project and is readable without authentication or decryption. It identifies a social-media account, follower information, an age and location profile, employment details, and characteristics of the user's computer.

Additional records disclose operational facts such as use of an OpenAI-compatible API and browser automation. No actual API secret or saved browser target identifier was found, so the confirmed issue is information disclosure rather than credential compromise.

The application also writes memory re ...[truncated 1098 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove all real user memories from the distributed tree.json.
  2. Replace the database with an empty schema or clearly synthetic demonstration records.
  3. Store runtime memory outside the installation directory, such as in a per-user application-data directory.
  4. Exclude runtime databases from source control and package publication through appropriate ignore and package-exclusion rules.
  5. Create files with owner-only permissions, such as mode 0600 on supported platforms.
  6. Document that credentials, session identifiers, private account data, and other secrets must not be stored as memory content.
  7. Add automated secret and personal-data scanning to the release process.
  8. If sensitive memories are an intended use case, provide encryption at rest and an explicit data-retention and deletion policy.

T08 · Insecure Dependencies

Note
Location
package.json:17
Finding

Unused and Non-Reproducibly Pinned Third-Party Dependency

Content
View full analysis

Vulnerability Details

File Location: package.json:17-19
Vulnerability Type: Unnecessary supply-chain exposure
Risk Level: Low

json
"dependencies": {
  "mem0ai": "^1.0.8"
}

Technical Analysis

The audited Python implementation does not import or invoke mem0ai, yet the package manifest declares it as a runtime dependency. The caret version constraint permits later compatible releases, and the project does not include a lockfile. Consequently, installation may resolve code that was not part of the reviewed artifact.

No evidence establishes that the current mem0ai package is malicious. The security issue is the unnecessary and mutable supply-chain attack surface created by installing an unused dependency and its transitive dependency graph.

Attack Path

  1. A user installs the project through an npm-compatible package manager.
  2. The package manager resolves mem0ai using the mutable ^1.0.8 range because no lockfile fixes the reviewed dependency graph.
  3. The package manager downloads the selected package and its transitive dependencies.
  4. If a future selected release or transitive dependency is compromised, malicious installation or runtime behavior could execute under the installing user's account.
  5. This exposure exists even though the audited Skill functionality does not require mem0ai.

This path is conditional on a dependency or dependency-distribution channel becoming compromised; no such compromise was confirmed during this audit.

Impact Assessment

A successfully compromised dependency could generally act with the privileges of the user running the package installation, subject to package-manager and operating-system restrictions. Potential scope includes access to that user's files, environment variables, network access, and modification of user-writable resources. The present artifact itself contains no confirmed dependency payload or direct exploitation code.

Remediation
View remediation

Remediation Suggestions

  1. Remove mem0ai from package.json because it is not used by mem0_skill.py.
  2. If it becomes necessary, pin an exact reviewed version rather than using a caret range.
  3. Commit a lockfile generated from a trusted registry to make dependency resolution reproducible.
  4. Review the dependency's provenance, maintainers, release history, transitive dependencies, and installation scripts before inclusion.
  5. Use automated dependency auditing and integrity verification in continuous integration.
  6. Install dependencies with lifecycle scripts disabled where they are not required.
  7. Keep the manifest aligned with the actual Python runtime and document why each third-party package is needed.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (7)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The file presents the skill’s operational description, usage instructions, and examples entirely in Chinese. Under the policy, forcing a specific language without giving the user a language or locale choice is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The memory system stores arbitrary user-provided content and later returns raw content, previews, dependencies, and related entries directly in responses, with no sensitivity filtering, access control, redaction, or tenant separation. If secrets, credentials, personal data, or internal notes are ever stored, a later query can disclose them to an unintended user or to a compromised agent flow.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill exposes a destructive clear() operation that immediately wipes the entire persisted memory tree and saves the empty state without any confirmation, authorization, rollback, or scope restriction. In an agent or multi-tool setting, an accidental invocation, prompt injection, or misuse by another component could permanently delete all stored data.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This JSON manifest-like file uses Chinese throughout its natural-language content fields and does not indicate that users can choose another language or locale. Under the language/locale policy rule, forcing a single language without opt-in is a policy concern when no justification or alternative is provided.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The module description and all user-facing CLI/help strings are written only in Chinese, which effectively imposes a specific language on users without offering a choice. Under the locale policy, language constraints should be optional or clearly justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The description string is written only in Chinese ("Mem0 深层记忆系统技能"), which signals a fixed language choice without offering any user language/locale option or documenting why this package is Chinese-only. Under the policy criteria, a skill that forces a specific language without opt-in can be a natural-language policy violation.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
89% confidence
Finding

The dependency version is specified with a caret range (^1.0.8), which permits automatic installation of newer compatible releases. This increases supply-chain risk because a compromised or vulnerable upstream release could be pulled in without explicit review, especially important for an agent skill that may process sensitive memory or user data.

Content

Scanner excerpt · package.json (reported line 18)May include surrounding context.

json
"author": "OpenClaw User",
  "license": "MIT",
  "dependencies": {
    "mem0ai": "^1.0.8"
  }
}