Back to skill

Security audit

Dream Journal

Security checks for vulnerabilities and agentic risk

Overview

This dream journal skill persistently stores dream entries and learned imagery locally as its documentation describes, with no evidence of hidden execution, credential access, network exfiltration, or destructive behavior.

Install only if you are comfortable having dream narratives, tags, learned imagery, and statistics saved in local memory files. Review or delete those files manually if needed, especially on shared or backed-up workspaces.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (8)

Tp4

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding

The description is broadly aligned with the core purpose: this is indeed a dream-recording and imagery-learning tool that stores dated entries, expands an imagery library, and provides search/analyze functions. It also generates a lightly literary title, which partially matches the '文学化修饰' claim. However, there are material mismatches in behavior versus description. Most importantly, the description says that when a user mentions dream-related content, the tool automatically extracts and records it, but the supplied code is just a command-line script requiring explicit invocation (record, search, analyze, library) and contains no trigger integration or automatic detection. In addition, some declared features are incomplete: search supports only a keyword through the CLI, while category and days are effectively unused; category filtering checks raw entry text rather than structured category metadata; and stats initialize category_counts but never populate them, so the analysis is narrower than implied. These are significant enough to mark as a description-behavior mismatch, even though the primary domain and most core functions do match.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
87% confidence
Finding

The skill clearly describes persistent reads/writes under the memory/ directory, but it does not declare any explicit tool scope or file access boundaries. That creates an unnecessary trust gap: the runtime may permit broader file operations than users or reviewers expect, increasing the risk of unintended access or modification beyond the dream journal data.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill handles highly personal dream narratives and also derives additional personal data through imagery extraction and long-term accumulation, yet the description does not clearly warn users that this data is automatically stored locally over time. That undermines informed consent and can expose sensitive psychological, emotional, or identifying information to anyone with local access to those files.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The code persists automatically learned user imagery and later statistics into JSON files, building a behavioral/profile dataset derived from sensitive user inputs without explicit disclosure or consent. Even when raw dream text is not directly reviewed, the accumulated tags and imagery can reveal recurring fears, people, places, trauma themes, religion, or health-related concerns.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The script writes full dream narratives, inferred tags, and learned imagery to persistent files under /root/.openclaw/workspace/memory without any consent prompt, retention control, or visibility to the user at write time. Dream journals often contain highly sensitive mental-health, relationship, sexual, or identity-related information, so silent persistence creates a real privacy and data-exposure risk if the workspace is shared, backed up, inspected by other tools, or later exfiltrated.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The script stores raw natural-language dream entries plus tags in a shared workspace memory location, which is exactly the kind of retained sensitive text that can be unintentionally exposed to other components, operators, backups, logs, or future prompts. In this skill's context, the content is especially privacy-sensitive because dream narratives can contain intimate personal thoughts and inferred psychological signals, making the retention risk more serious than ordinary note-taking metadata.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The visible natural-language content is entirely in Chinese and provides no indication that users can choose another language or locale. Per the policy, forcing a specific language without opt-in is a natural-language policy concern unless it is clearly documented as region-specific or optional.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The natural-language description and command-line messages are entirely in Chinese, which imposes a language constraint without any opt-in or documented reason. Under the policy, forcing a specific language without user choice can be a locale-policy issue unless the restriction is clearly justified.

Content

No source excerpt is available for this finding.