Back to skill

Security audit

Openclaw Cc Contrib

Security checks for vulnerabilities and agentic risk

Overview

This skill is not proven malicious, but it automatically saves, rewrites, and can delete conversation memories in the background, so it needs careful review before installation.

Install only if you want OpenClaw to keep long-term notes from your conversations. Before enabling it, disable or avoid the cron job unless you really want background consolidation, review proposed memories before saving when possible, keep backups of memory files, and avoid using it in chats that include secrets, credentials, private URLs, or sensitive account details.

Vulnerability Patterns
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • System PersistenceInstalls backdoors, hooks, services, or scheduled tasks that survive the run
  • 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
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Findings (5)

T06 · System Persistence

Error
Location
dream-rem/SKILL.md:161
Finding

Recurring Cross-Session Scheduled Task Creates System Persistence

Content
View full analysis
Remediation
View remediation

T02 · Agent Memory Poisoning

Error
Location
extract-memories/SKILL.md:23
Finding

Untrusted Conversation Content Can Be Persisted as Long-Term Agent Guidance

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
extract-memories/SKILL.md:42
Finding

Unbounded Session Collection Can Persist Secrets and Sensitive Identifiers in Plaintext

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
dream-rem/SKILL.md:42
Finding

Scheduled Consolidation Performs Destructive Memory Changes Without Approval or Rollback

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Note
Location
UPDATE_FLOW.md:53
Finding

Unsafe Temporary-Directory Update Procedure Can Strand Runtime Skills

Content
View full analysis
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 (18)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill description states that conversation information will be written to the memory/ directory for persistent reuse, but it does not provide a clear warning or informed consent flow to the user. Because the data being extracted includes decisions, preferences, project conventions, and technical resolutions from full conversation context, this creates a meaningful privacy and data-retention risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill states that it will proactively analyze the conversation and write durable memory entries into memory/topics/ and MEMORY.md, but it does not require clear user notice or consent at the moment of retention. Persisting user decisions, preferences, constraints, and external references without transparent warning creates privacy, compliance, and trust risks, especially if the stored data later becomes accessible beyond the immediate session.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
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.

Content

Scanner excerpt · README.md (reported line 44)May include surrounding context.

📦 安装

bash
mkdir -p ~/.openclaw/skills/
cp -r extract-memories ~/.openclaw/skills/
cp -r memory-sorting ~/.openclaw/skills/
cp -r dream-rem ~/.openclaw/skills/

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The README states that the skill will automatically extract conversation-derived memories and write them into persistent topic files, but it does not prominently warn users that their chat content may be stored beyond the session. In a memory-management skill, silent persistence of user-provided information creates privacy and consent risk, especially because the trigger is automatic rather than explicitly user-confirmed each time.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The README describes deep consolidation steps that can delete outdated content and rewrite MEMORY.md, but it does not clearly warn that maintenance operations may remove or overwrite existing data. That creates integrity and recoverability risk because users may assume the tool is additive when it is actually destructive during cleanup.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
81% confidence
Finding

The manifest description and trigger phrases are presented in Chinese, and the skill behavior implies fixed-language interaction for activation and notification without documenting any user language choice. This can violate language or locale policy when a skill enforces a specific language without opt-in or clear region-specific justification.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill explicitly instructs the agent to read the full conversation, extract user preferences, decisions, conventions, and solutions, and persist them to long-term storage automatically. In context, this is more dangerous because the extraction happens after every conversation by default, so sensitive personal, organizational, or project data may be retained beyond the original interaction without granular review.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill is configured to run automatically at the end of every conversation, which is an overly broad trigger for an operation that reads the full session and writes persistent memory. This can cause unintended invocation on sensitive chats, increasing the chance that private or irrelevant information is stored without an explicit user action at that moment.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

This markdown file is written entirely in Chinese and presents the process as a mandatory standard, but it does not indicate that language is optional or limited to a China-specific audience. Under the policy criteria, forcing a specific language without user opt-in can be a natural-language policy violation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly directs deletion, rewriting, and pruning of memory artifacts, including removing 'outdated' or 'incorrect' content, but it does not require a clear user-facing warning, confirmation, backup, or dry-run before destructive changes. In a memory-management skill, this is especially dangerous because normal operation can irreversibly erase user data, contradictions, or historical context based on potentially incorrect model judgments.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The core behavior of the skill is to extract conversation-derived user information and external references and save them into persistent files. Even if intended as a memory feature, automatically converting session content into durable records can capture sensitive personal or project information that was only meant for transient conversational use.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The automatic end-of-conversation and heartbeat triggers instruct the agent to read session history and persist derived content without a per-instance user request. This makes the data collection less context-aware and more likely to store unintended or sensitive material, especially because heartbeat and farewell heuristics can fire when the user did not explicitly ask to save anything.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill defines broad automatic end-of-conversation triggers such as common farewell phrases and also permits heartbeat-based invocation, which can cause memory extraction during ordinary chat flow rather than only after an intentional user action. Because the skill performs persistence of conversation-derived data, overbroad triggering increases the chance of collecting and storing information the user did not expect to retain.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The manifest description and main instructions are presented in Chinese, and the listed trigger phrases are also Chinese-centric except for the command name. This can impose a language requirement on users without opt-in, which matches the policy concern for forced language or locale usage.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The entire skill description, including trigger phrases and output expectations, is written only in Chinese and does not offer an alternative language or indicate that Chinese is an optional locale. This can violate language/locale policy when a skill implicitly requires a specific language without user opt-in or documented regional justification.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill declares an optional automatic trigger after code submission without defining narrow activation criteria, scope limits, or user confirmation. In an agent environment, broad auto-triggering can cause unintended execution on unrelated repositories or sensitive code changes, leading to surprise data access, unnecessary analysis, or workflow disruption.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

All user-facing descriptive content and trigger examples in this file are presented in Chinese, with no indication that other languages are supported or that the language choice is optional. Under the policy, forcing a specific language without opt-in can be a natural-language policy violation unless clearly justified and documented.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The manifest description, headings, triggers, workflow, and command guidance are written predominantly in Chinese, with no indication that another language is supported or that the user can opt into this locale. Under the stated policy, forcing a specific language without user choice is a natural-language policy concern.

Content

No source excerpt is available for this finding.