Back to skill

Security audit

VBot Skills Framework

Security checks for vulnerabilities and agentic risk

Overview

The skill bundle is mostly documentation and examples, but it combines broad activation rules with network lookup, scheduled reminder, GitHub API, and dependency-update workflows without clear scoping or user-control guidance.

Install only if you want a broad multi-skill bundle. Review or narrow the triggers before use, and require explicit confirmation before web searches, GitHub API calls, npm updates, reading diffs, or setting recurring reminders. Treat the malformed package.json as a packaging quality issue to fix before relying on automated installation tooling.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Rogue AgentSelf-Modification, Session Persistence
  • 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
Findings (19)

Self-Modification

High
Category
Rogue Agent
Confidence
85% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · README.md (reported line 18)May include surrounding context.

bash
cp -r template my-skill
cd my-skill
# Edit SKILL.md

2. Define Your Skill

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The sample manifest uses the description "When user wants to do X," which is too generic to convey clear activation boundaries. As documentation for how to define a skill, it models an ambiguous trigger style that can overlap with many user intents and does not include scope constraints or negative examples.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The README suggests using a one-line description of "When to use" and makes triggers optional, but it does not require specificity or constraints. In a framework document that teaches skill authors how to define invocation conditions, this can lead to ambiguous activation conditions and broad trigger overlap.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The activation guidance is overly broad and delegates activation to any situation where a user's request merely resembles the skill description, without explicit triggers, scope boundaries, or authorization checks. In an agent framework, this can cause unintended skill invocation and broaden the attack surface for prompt injection, unsafe tool use, or execution of capabilities the user did not clearly request.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

This is a manifest file, so vague-trigger checks apply. The package metadata and clawhub block describe the skill as a general-purpose framework for reusable AI agent skills, but provide no specific activation phrases, scope limits, or exclusion conditions, leaving invocation conditions ambiguous.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger list includes very short and generic terms such as "dep" and "outdated", which can cause the skill to activate in unrelated contexts. While this skill only suggests dependency-checking commands and does not itself contain dangerous actions, unintended activation can still lead to user confusion or accidental execution of package-management workflows.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger list includes very broad terms such as "emoji" and especially "icon", which can cause the skill to activate in unrelated contexts. This can lead to unintended routing or skill selection errors, reducing reliability and potentially interfering with more appropriate skills, though it does not directly introduce code execution or data exposure risk.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The activation guidance, "Activate when user asks about GitHub repos," is broad enough that an agent may invoke this skill for many loosely related repository questions without clear scoping or guardrails. Overly broad triggers can cause unintended skill execution, increasing the chance of unnecessary external lookups, incorrect tool selection, or disclosure of context to the wrong workflow.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger list includes very broad terms such as "github" and "repo", which can cause this skill to activate in many unrelated contexts. Over-broad activation can misroute user requests, cause unintended tool/API usage, and increase the chance that a different skill is bypassed or overshadowed.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The phrase "Activate when user wants to set a reminder" is broad and does not define specific trigger phrases, scope limits, or exclusion conditions. This could cause unintended invocation because many ordinary requests may implicitly involve reminders without clearly signaling that this skill should activate.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The trigger list includes Chinese terms alongside English, but the example triggers and phrasing strongly center Chinese usage without stating that users may choose their preferred language or locale. Under the policy, language constraints should be opt-in or clearly justified, and this file does not document such a choice.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The activation guidance 'Activate when user wants to search or lookup information' is very broad and can match many ordinary requests, causing the skill to trigger in situations where a narrower or more appropriate capability should be used. Over-broad activation increases the chance of unnecessary web access, unintended data exposure to external services, and incorrect agent behavior driven by ambiguous routing.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger list is broad and contains generic terms like 'search', '搜索', and '查找', which are likely to appear in ordinary conversation and can cause unintended invocation of the web-search skill. Unintended routing to a network-enabled skill can expand the assistant’s permissions or behavior surface, potentially causing unnecessary external lookups, privacy issues, or bypass of more appropriate local-only handling.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The example trigger uses ambiguous everyday phrasing such as '帮我查一下这个...', which could match many normal requests that do not necessarily require web access. This increases the chance that the skill is selected when the user intended a different capability, leading to over-broad network use and less predictable agent behavior.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
60% confidence
Finding

Dependencies lack version pinning, allowing potential malicious package updates. Consider pinning versions.

Content

Scanner excerpt · package.json (reported line 13)May include surrounding context.

json
"slug": "vbot-skills-framework",
    "name": "VBot Skills Framework"
  }
EOF

# 重新发布
clawhub publish ./ --slug vbot-skills-framework --name "VBot Skills Framework" --version 1.0.0 --changelog "Initial release" 2>&1

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

Natural-language policy checks apply to all file types. The standalone Chinese text '重新发布' introduces a specific language without offering user choice or documenting that the skill is intended for a Chinese-language or region-specific context.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
76% confidence
Finding

The file includes both English and Chinese trigger/examples (for example, "仓库" and "这个仓库有多少 stars") but does not state whether responses should follow the user's language preference or how locale is selected. This can create an implicit language policy without explicit user opt-in or documented behavior.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The description says the skill sets reminders and scheduled notifications, but it does not warn users that it may create notifications or timed actions that persist beyond the current interaction. For a markdown skill description, this omits a potentially relevant disclosure about behavior that can affect the user's environment or workflow.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

This markdown/manifest template shows example activation text like "Use when user wants to do X" and triggers such as "do X" and "make X happen," which are intentionally generic and could encourage authors to define vague invocation conditions. Because the template does not pair these examples with exclusion guidance or specificity requirements, it risks propagating ambiguous trigger design into derived skills.

Content

No source excerpt is available for this finding.