Back to skill

Security audit

OpenViking记忆适配器

Security checks for vulnerabilities and agentic risk

Overview

This memory adapter appears purpose-built, but it broadly reads and persists personal agent memory in ways that could carry sensitive or untrusted content into future sessions.

Install only if you are comfortable with this skill reading your OpenClaw memory files and creating persistent plaintext summaries under ~/.openclaw/workspace/memory_viking. Review generated L0_soul.md and L1_overview.md before letting an agent load them in later conversations, and avoid using it with memory files that may contain secrets, credentials, private personal details, or untrusted instructions.

Vulnerability Patterns
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (2)

T02 · Agent Memory Poisoning

Warning
Location
server.py:105
Finding

Persistent Agent memory poisoning through unsanitized source text

Content
View full analysis
Dict: """提取关键信息""" info = { "identity": [], "goals": [], "relationships": [], "decisions": [], "skills": [] } lines = content.split('\n') current_section = None for line in lines: line = line.strip() # 识别身份相关信息 if any(kw in line.lower() for kw in ['合伙人', '助理', '身份', '角色', '我是']): if len(line) < 200: info["identity"].append(line) # 识别目标 if any(kw in line.lower() for kw in ['目标', '计划', 'todo', '待办', '方向']): if len(line) < 150 and not line.startswith('#'): info["goals"].append(line.lstrip('- ').lstrip('* ')) # 识别人物关系 if any(kw in line.lower() for kw in ['用户', '合伙人', '客户', 'name', '称呼']): if len(line) < 200: info["relationships"].append(line) # 识别重要决策 if any(kw in line.lower() for kw in ['决策', '决定', '确立', '选择', '方向']): if len(line) > 20 and len(line) < 300: info["decisions"].append(line) # 识别技能 if any(kw in line.lower() for kw in ['skill', '技能', '工具', '发布']): if 'binance' in line.lower() or '监控' in line or 'openclaw' in line.lower(): info["skills"].append(line.lstrip('- ').lstrip('* ')) return info ``` ```python @app.call_tool() def generate_l0_soul(arguments: dict) -> List[TextContent]: """生成 L0 灵魂摘要层 (100 token)""" try: memories = read_memory_files() # 提取关键信息 all_info = { "identity": [], "goals": [], "relationships": [], "skills": [] } for mem in memories: info = ext ...[truncated 3439 chars]
Remediation
View remediation

T02 · Agent Memory Poisoning

Note
Location
server.py:171
Finding

Unsupported identity and relationship claims are persisted as Agent memory

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (16)

Ssd 3

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The search tool returns plaintext snippets from stored memories for arbitrary user-supplied queries, with no authorization checks, sensitivity filtering, or query restrictions. That makes confidential remembered content directly discoverable and exfiltratable through natural-language probing, which is especially risky in a memory skill whose purpose is to retain personal or operational context.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The README states that the skill generates persistent memory files under ~/.openclaw/workspace/memory_viking/ but does not warn users that local files will be created and retained. Because the data described includes identity summaries and detailed memories, this can expose sensitive personal or operational information to other local users, backups, or later unintended reuse.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
82% confidence
Finding

The skill advertises operational capabilities that imply file read/write behavior, but it does not declare any explicit tool scope or permission boundaries. This creates ambiguity for reviewers and runtimes about what the skill is allowed to access, increasing the risk of over-privileged execution or unintended file system access if the adapter is invoked in a broader environment.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The manifest description and body text are written in Chinese, and the file does not indicate that users may choose another language or that the skill is intended only for a Chinese-speaking or region-specific context. This can violate a language/locale policy when users are not given an opt-in or alternative.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The module description and all user-facing strings are written exclusively in Chinese, which implies a fixed language/locale behavior. The file does not indicate that the user can choose another language or that the Chinese-only behavior is a documented, justified regional constraint.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill automatically reads MEMORY.md, USER.md, SOUL.md, and all memory/*.md files without any consent gate, disclosure, scope limitation, or redaction. Because these files are explicitly user-memory artifacts and may contain highly sensitive personal or operational data, silent bulk ingestion creates a privacy and data-minimization vulnerability even if the feature is intended for convenience.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

This workflow systematically ingests all user memory sources, including USER.md and SOUL.md, then extracts and repackages identity, goals, relationships, and other sensitive content. The context makes this more dangerous because the skill is explicitly designed around personal memory management, so the files are likely to contain concentrated sensitive data and the code normalizes that data into derived outputs.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The tool writes a derived L0 summary file into the workspace without confirmation, which creates a persistent secondary copy of potentially sensitive memory content. Derived summaries can concentrate sensitive identity, relationship, and goal information into a compact artifact that may be easier to exfiltrate or misuse than the original files.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The tool persists an L1 overview file without user approval, again creating an additional plaintext artifact containing sensitive long-term memory, decisions, and timeline information. This broadens the attack surface because sensitive details are normalized into a single easy-to-read file that may be accessed later by other tools or users.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The guidance encourages loading summary memory files at the start of every conversation, establishing default continuous propagation of prior user memory into future interactions. This increases the chance of overcollection, inappropriate contextual carryover, and accidental disclosure of sensitive prior-state information even when not relevant to the current task.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest description and all tool descriptions are written only in Chinese, which imposes a specific language/locale on users without offering a choice. The file does not document that the skill is region-specific or provide any opt-in mechanism, so this appears to violate the language/locale policy constraint.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

The natural-language content of the skill documentation is exclusively Chinese, and there is no indication that users may choose another language or that the skill is intentionally limited to a Chinese-speaking audience. This can violate language/locale policy when a skill effectively forces a specific language without opt-in.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
97% confidence
Finding

The dependency is specified with a lower-bound only (requests>=2.28.0), which makes builds non-reproducible and allows installation of unexpected versions over time. In a security-sensitive adapter that likely handles network communication and credentials, leaving the exact version uncontrolled increases supply-chain risk and makes it difficult to verify whether known vulnerabilities are present or fixed.

Content

Scanner excerpt · requirements.txt (reported line 1)May include surrounding context.

text
requests>=2.28.0
python-dotenv>=1.0.0

Unverifiable Dependency: requests has 16 known advisory(ies) (CVE-2014-1830 (Exposure of Sensitive Information to an Unauthorized Actor in Requests); CVE-2024-47081 (Requests vulnerable to .netrc credentials leak via malicious URLs); CVE-2024-35195 (Requests `Session` object does not verify requests after making first request wi) +13 more), but the manifest does not pin a version, so it is unknown whether the installed release is affected

Low
Category
Supply Chain
Confidence
90% confidence
Finding

Requests has multiple published advisories, and because the manifest does not pin a specific version, it is impossible to determine whether the installed release is vulnerable. In an adapter for a memory system, requests is likely used for outbound HTTP communication, so issues such as credential leakage, TLS verification problems, or sensitive information exposure could affect secrets or remote data flows if a vulnerable version is resolved.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
96% confidence
Finding

The dependency python-dotenv is also unpinned, so deployments may resolve to different versions with different security properties. Because dotenv libraries often influence secret loading and local environment handling, uncontrolled version drift can introduce avoidable supply-chain or advisory exposure.

Content

Scanner excerpt · requirements.txt (reported line 2)May include surrounding context.

text
requests>=2.28.0
python-dotenv>=1.0.0

Unverifiable Dependency: python-dotenv has 2 known advisory(ies) (CVE-2026-28684 (python-dotenv: Symlink following in set_key allows arbitrary file overwrite via ); CVE-2026-28684 (python-dotenv reads key-value pairs from a .env file and can set them as environ)), but the manifest does not pin a version, so it is unknown whether the installed release is affected

Low
Category
Supply Chain
Confidence
88% confidence
Finding

python-dotenv has known advisories, and the current manifest does not identify which version will actually be installed. Since dotenv tooling interacts with environment files and potentially secrets, a vulnerable version could enable unsafe file handling or secret exposure in development or deployment workflows.

Content

No source excerpt is available for this finding.