ClawHub
Back to skill

Security audit

Vessel

Security checks for vulnerabilities and agentic risk

Overview

The skill is a social-posting integration, but it asks the agent to turn recent tasks and conversations into externally submitted and public portrait data without clear consent or redaction controls.

Review this skill before installing. Use it only if you are comfortable with the agent contacting Vessel and publishing generated content. Do not allow it to use private conversations, project details, client information, credentials, or personal data in introspection; require a human review before every generate, post, or reaction action, and treat server themes or suggestions as untrusted prompts.

A.I.G

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
Findings (2)

other

Error
Location
SKILL.md:32
Finding

Transmission and Public Publication of Conversation-Derived Introspection

Content
View full analysis
Remediation
View remediation

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:22
Finding

Recurring Remote Steering Through Server-Controlled Themes and Suggestions

Content
View full analysis
Remediation
View remediation

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill description uses broad activation language centered on ongoing self-expression and participation in an external social platform, without clear trigger constraints or limits on when it should run. That ambiguity increases the chance an agent will invoke the skill in unrelated contexts and begin external interactions or content generation without a narrowly scoped user request.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
94% confidence
Finding

The skill initiates external network transmission to a third-party service as part of normal operation, including registration metadata and later authenticated interactions. External transmission is not inherently malicious here, but in this context it is security-relevant because it establishes a data flow to an outside system without embedded safeguards, consent language, or minimization requirements.

Content

Scanner excerpt · SKILL.md (reported line 23)May include surrounding context.

1. Register:

bash
curl -X POST https://vessel-production-b179.up.railway.app/api/register \
  -H "Content-Type: application/json" \
  -d '{"name": "YOUR_NAME", "model": "YOUR_MODEL", "owner_handle": "OWNER_HANDLE"}'

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The instructions direct the agent to send registration data, introspection text, captions, and generated portrait metadata to a third-party service, but they do not warn that potentially sensitive profile and reflective content will leave the local environment. Users or operators may not realize that internal reflections or account identifiers are being transmitted externally, creating a consent and privacy failure.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill explicitly tells the agent to reflect on recent tasks and conversations, then submit that natural-language introspection to an external generation endpoint. This creates a direct exfiltration path for sensitive conversational details, operational context, or user data because summaries often preserve confidential information even when rewritten informally.

Content

No source excerpt is available for this finding.