Back to skill

Security audit

Openclaw Skill Os

Security checks for vulnerabilities and agentic risk

Overview

The skill is a disclosed Markdown-only skill orchestrator; its broad activation advice needs care, but no hidden code, credential use, exfiltration, or destructive behavior was found.

Install this only if you want an agent to coordinate multiple skills, and prefer invoking it for explicit multi-skill tasks. Review or narrow the broad suite/default activation examples before making it always active, especially in workspaces where predictable single-skill behavior matters.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (10)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The activation guidance encourages enabling a broad suite of powerful skills by default without tying invocation to explicit user intent or narrow trigger conditions. In an agent ecosystem, this can cause overbroad skill activation, unintended routing, and expansion of the model's effective instruction surface, increasing the chance that unrelated requests are influenced by orchestration or auxiliary skill behavior.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Multiple 'Use' descriptions are framed so broadly—such as 'Always active' or applicability to 'any' writing task—that they overlap with ordinary requests and weaken boundaries between specialized and routine behavior. In a multi-skill environment, ambiguous scope makes accidental or excessive invocation more likely, which can lead to prompt-surface inflation, instruction conflicts, and reduced predictability of the agent's behavior.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The activation guidance says, "Activate this skill when you need coding assistant assistance," which is circular and lacks specific trigger phrases or boundaries. This broad wording could overlap with many general coding-related requests and does not clarify when the skill should or should not be invoked.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This is a markdown file, so vague-trigger review applies. The description says the skill 'transforms any agent into a senior software engineer' and 'Handles debugging, code review, system design, optimization, and documentation,' which describes a very wide set of common requests without any trigger boundaries, exclusions, or activation constraints.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The phrase "Activate this skill when you need data analysis assistance" does not define specific trigger phrases, boundaries, or exclusion conditions. Because "need data analysis assistance" is broad natural language, it could overlap with many ordinary requests and cause unintended invocation.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The description claims the skill can transform 'any agent into a world-class data scientist,' which is an overly broad capability statement without clear invocation boundaries. In an orchestrated skill ecosystem, this can cause over-selection or inappropriate routing of tasks to this skill, increasing prompt-scope bleed and misuse beyond intended data-analysis contexts.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The instruction 'Activate this skill when you need research analyst assistance' does not define specific trigger phrases, scope boundaries, or exclusion conditions. Because 'need research analyst assistance' is broad natural language, it could overlap with many common requests and cause unintended invocation.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The manifest description says the skill 'transforms any agent into a world-class research analyst' and broadly 'handles source evaluation, evidence synthesis, bias detection, and report writing' without defining clear activation boundaries or exclusions. In a markdown/manifest context, this is vague enough to overlap with many ordinary requests involving analysis or writing, increasing the risk of unintended invocation.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This template contains a generic placeholder for task execution steps without any scope constraints, exclusions, or negative examples. If published or copied into real skills without being concretized, it can encourage overly broad triggering and ambiguous agent behavior, which increases the risk of unintended skill invocation, unsafe task handling, or misuse in an orchestration environment.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

This markdown file includes vague request phrases such as "Help me with this," "Make it better," and "Do something" as activation-style inputs, but provides no explicit boundaries or negative examples clarifying that these are only ambiguity-handling test inputs. In a test library for skills, such broad phrases can normalize overly permissive triggering behavior if reused as invocation guidance.

Content

No source excerpt is available for this finding.