Back to skill

Security audit

autopilot-loop

Security checks for vulnerabilities and agentic risk

Overview

The scanners classify this skill as benign with medium confidence: they found no AIG issues and only two contextual Skillspector flags, though reference checks were partial.

The verdict is benign, but confidence is medium. Consider whether the autonomy noted in the scan fits your needs. Coverage is incomplete: both listed Markdown components were scanned, but references to local paths could not be resolved, so those referenced materials were not assessed.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 27)May include surrounding context.

md
If `goal.md` does not exist, interview the owner one question at a time:
1. What is the goal, with a number and a deadline?
2. What niche or project is this for?
3. What may you do without asking? What always needs approval?
4. What tools, accounts, and files may you use?
5. What has been tried before, and where did it stall?
6. What name goes on messages, and through what channel do they go out (email, text, post)?

Generic invocation phrases can overlap with ordinary planning and review requests.

Low
Category
Trigger Abuse
Confidence
90% confidence
Finding

“Plan the week,” “Daily run,” and “Evening review” are presented as invocation phrases without an Autopilot-specific qualifier. These phrases could refer to unrelated planning or review tasks, and the README does not explicitly restrict them to an already initialized Autopilot session.

Content

No source excerpt is available for this finding.