Back to skill

Security audit

Openclaw Brain Core Ultra

Security checks for vulnerabilities and agentic risk

Overview

This skill is a read-only reasoning framework with broad but disclosed scope, and the current runtime file includes limits for trivial, unsafe, or out-of-scope tasks.

Before installing, understand that this is a broad reasoning-style skill, so it may make the agent more structured and verification-heavy on complex tasks. It should be kept scoped to difficult, risky, or multi-step work, and memory/observability behavior should remain under the platform's normal controls.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (8)

Vague Triggers

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The usage guidance explicitly says to activate the framework on every task, which directly conflicts with the stated limited purpose. In context, this is more dangerous because the skill is a meta-framework controlling reasoning, tools, verification, and other behavior; universal activation would let it persistently shape nearly all agent actions.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The README instructs users to apply the skill for broad, subjective conditions like 'high-level reasoning' and 'task kompleks berdependensi banyak' without concrete activation criteria, exclusions, or safeguards. Because this is a meta-skill that 'shapes every response,' ambiguous scope can cause over-application, policy interference, or unintended delegation of decision-making logic across many tasks, increasing the chance of unsafe behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The frontmatter and compliance wrapper present the file as changelog/usage guidance, but the body actually defines a broad agent-governing framework with decision, execution, memory, and verification policies. This misrepresents the skill’s effective authority and can bypass user or platform expectations about what the skill is allowed to do, increasing the chance of unintended control over agent behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The skill defines its identity and purpose in Indonesian while other sections are in English, and the instructions later direct activation on broad task use without stating that language should follow user preference. This can amount to a language-policy issue because the skill implicitly imposes a mixed/default locale rather than explicitly offering opt-in or user choice.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger condition activates whenever the user request 'matches the description,' which is broad and subjective. For a meta-skill that governs reasoning and action selection, vague activation criteria make it easier for the skill to attach to unrelated tasks and exert unintended control.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest says this skill is for deep reasoning on complex tasks, but the content asserts a meta-skill that shapes every response and later instructs activation on every task. That scope expansion makes the skill effectively omnipresent, allowing it to override narrower routing and influence low-risk tasks that never needed this level of authority.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

The skill introduces memory retention and update behavior even though its stated purpose is reasoning guidance. Unnecessary memory directives can cause over-collection, persistence of sensitive task context, or cross-task influence without clear user need or authorization.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The skill mandates observability and event emission despite being presented as a reasoning framework. Even with a note to exclude secrets, unnecessary telemetry requirements expand the skill’s operational scope and can increase exposure of metadata, task traces, or sensitive context in logs.

Content

No source excerpt is available for this finding.