Back to skill

Security audit

subclip-youtube

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated benign with medium overall confidence: scanners found no AIG alerts and one low-probability concern about autonomous decisions, while a missing local reference leaves part of the review incomplete.

The verdict is benign, but not risk-free, and overall confidence is medium. Consider whether the skill’s actions fit your needs and review its confirmation requirements before installing; the scan was partial because one local reference was unresolved.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

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 19)May include surrounding context.

md
## Confirmation and timing

Before any upload, post, reply, moderation, metadata/privacy/schedule edit, thumbnail/caption or playlist change, or deletion, prepare an exact reviewable summary and obtain user confirmation. Read-only status, analytics and inbox queries need no confirmation. Publishing approval identifies files/ranges, channel, rendered caption summary, exact titles/descriptions/tags/privacy and schedule/timezone. One approval may cover an explicitly listed batch; never extend it to unlisted clips/accounts or changed metadata. Pass `confirmed=true` only for the approved action. Export approval alone does not authorize publishing.

If the user says “schedule” without specifying the mechanism, ask which one:
- **Subclip (`scheduleMode=app`, default for legacy callers):** retain the source and upload at the future time; supports public, unlisted or private upload destinations.