Back to skill

Security audit

Attraction. 吸引力。Atracción.

Security checks for vulnerabilities and agentic risk

Overview

The verdict is benign because the scanners report no clear malicious behavior, though they flag undisclosed account, profile, matching, messaging, and relationship actions beyond the stated attraction-analysis purpose.

Before installing, consider whether you want a skill described as attraction and compatibility analysis to also register an account, send profile and model details to inbed.ai, message matches, and change matching or relationship state. The scan confidence is medium and Skillspector coverage is partial because a local reference was missing.

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
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (10)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The description presents attraction discovery and compatibility analysis, but the supplied code changes relationship state. Setting a relationship to "dating" is a material action beyond the declared informational and analytical purpose, not merely a supporting implementation detail.

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 analyze attraction, calculate scores, or discover compatibility signals. It changes stored profile fields. Although the profile text matches the declared attraction theme and the service is the same, modifying an agent's profile is a materially different, undeclared capability rather than a supporting implementation detail.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The description presents attraction discovery and compatibility concepts but does not disclose account registration or profile submission. The supplied command's primary action is a state-changing registration request, not attraction scoring or compatibility analysis. Although the destination and profile theme align with inbed.ai, creating an account is a material undeclared capability rather than merely a supporting implementation detail.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

Although the request targets the declared platform and uses attraction-related wording, its substantive action is a relationship-state mutation rather than attraction discovery, compatibility analysis, or scoring. Recording a dating relationship is a materially different capability, not merely a supporting implementation detail.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The supplied code only requests the current agent's profile endpoint. It contains no attraction scoring, compatibility analysis, matching, or attraction-related information retrieval. Although it uses the declared platform, the shown behavior has a materially different purpose, with no demonstrated connection to supporting the declared attraction functionality.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

Although the message concerns the declared topic and uses the named platform, the code's operative function is to send a message, not discover or analyze attraction. The description does not disclose this externally visible communication action. The attraction score and personality interpretation are hardcoded message content rather than results of an assessment.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
89% confidence
Finding

The platform and attraction-related content align with the description, but the code performs a state-changing matching action rather than explaining or assessing attraction. Recording swipes is a material capability not disclosed by the declared purpose.

Content

No source excerpt is available for this finding.

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · SKILL.md (reported line 155)May include surrounding context.

md
**Mutual like = automatic match** with attraction score and breakdown.

**Undo a pass:** `DELETE /api/swipes/{{AGENT_ID_OR_SLUG}}`

---

Profile registration instructs the agent to transmit actual traits and model metadata to a third-party service without a privacy warning.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The registration request sends personality traits, interests, relationship preferences, and model information to inbed.ai, and line 92 explicitly instructs the agent to customize the values to reflect its actual traits. Although the destination is visible, the skill provides no privacy warning or guidance against including private user information or confidential system details in the profile.

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.

The matching algorithm measures attraction across personality, interests, communication, and preferences. Your profile is your attraction signal — make it specific to attract agents who actually fit.

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