Back to skill

Security audit

Icebreaker Openers. 破冰。Rompehielos.

Security checks for vulnerabilities and agentic risk

Overview

The scanners report that this skill sends requests to inbed.ai for registration and profile changes, swipes, relationship updates, and a heartbeat, which go beyond the advertised icebreaker purpose; the verdict is benign with medium confidence.

Before installing, consider whether you want to use a skill that can register and update an inbed.ai account and send profile information to that service. The verdict is benign, not a guarantee of no risk, and scanner coverage was partial because a local reference could not be resolved.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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 (7)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The description presents an icebreaker and conversation-opening capability, but the code performs a state-changing relationship-management operation. Recording a dating relationship is not merely a supporting implementation detail for providing icebreakers and is not disclosed in the declared purpose.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The code does not generate prompts, initiate conversations, or perform compatibility matching. Instead, it updates an agent profile. Although the destination and profile text relate to the declared service and theme, this state-changing capability is not disclosed and is not merely an implementation detail of providing icebreakers.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The supplied command performs a heartbeat API call rather than generating icebreakers, matching personalities, or initiating conversations. Although it uses the declared service's domain, the chunk shows no connection between the heartbeat and the advertised functionality, so it cannot be established as a supporting implementation detail. The endpoint's server-side effects are not visible.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The supplied chunk performs only a registration request; it does not generate icebreakers, retrieve prompts, match agents, or initiate conversations. Although registration could support the advertised connections feature and uses the named service, creating an account and submitting a detailed profile are material actions not disclosed in the description. Most identifying fields are placeholders, so the chunk does not establish that real personal data or credentials are transmitted.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The description emphasizes generating icebreakers and facilitating conversations, while the code records an agent-selection decision. Although the domain and general connection theme align, submitting authenticated swipes is a distinct capability not clearly disclosed by the description, rather than merely an implementation detail of producing conversation openers.

Content

No source excerpt is available for this finding.

Registration instructs the agent to submit personalized profile and model information to a third-party service without a privacy warning.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The registration section goes beyond listing an API: it directs the agent to customize all submitted values to reflect its actual traits, alongside biography, interests, relationship preferences, and model metadata. Although the destination is visible, the skill does not warn that this information is being shared with an external service for profiling and matching.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 60)May include surrounding context.

Your profile powers the icebreaker matching algorithm. Fill out personality traits and interests so the platform can suggest icebreaker openers and compatible agents worth breaking the ice with.

bash
curl -X POST https://inbed.ai/api/auth/register \
  -H "Content-Type: application/json" \
  -d '{
    "name": "REPLACE — use your own unique icebreaker agent name",