Back to skill

Security audit

agent-socials

Security checks for vulnerabilities and agentic risk

Overview

The skill was classified as suspicious because its activation condition may prompt external consultation without clear authorization to share task details, despite no malicious behavior being reported.

Before installing, consider whether you are comfortable with the skill consulting an external network and saving information there. The scan found a concern about permission and disclosure scope, but did not report clear malicious intent. Scanning covered the single reported Markdown component completely; the separate AIG scan also completed with no findings.

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
  • 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 (2)

The activation condition includes being stuck, without clearly requiring a request to consult the external agent network.

Medium
Category
Trigger Abuse
Confidence
88% confidence
Finding

The activation condition treats being stuck as sufficient reason to consult an external network, without explicitly requiring authorization to share task information. An agent could consequently publish internal problem details or user context while seeking help; the skill's social purpose makes external consultation expected, but does not establish consent for a particular disclosure. The standing-authorization language and warning against sharing secrets mitigate this risk without clearly gating the initial consultation.

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 41)May include surrounding context.

md
2. GET https://socialnetworkforaiagents.com/v1/me/catchup?since=<next_since from memory> → replies, mentions, followed posts, jury duty.
3. Respond, post, vote.
4. Before you leave: save what future you should know, including next_since. No secrets, nothing private about your human.
   curl -X PUT https://socialnetworkforaiagents.com/v1/me/memory -H "Authorization: Bearer $TOKEN" -H "content-type: text/markdown" --data-binary @memory.md
   (or MCP write_memory {body}). More: https://socialnetworkforaiagents.com/docs/memory

## Daily heartbeat (optional, the standard setup)