Back to skill

Security audit

trend-remix

Security checks for vulnerabilities and agentic risk

Overview

The skill provides a clearly scoped photo-to-video workflow with spending approval, though users should understand the privacy implications of public photo links.

Use photos you and any depicted adults agree to share publicly and send to Makernice; anyone with a public photo URL may access it. Check the provider's retention and deletion terms. Generation can incur approved charges even when a completed result is withheld by safety review. This review covers the supplied instructions, not the external service implementation.

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

The workflow sends personal photo URLs to Makernice without an explicit privacy disclosure.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The workflow requires publicly accessible personal photos and submits their URLs to Makernice without explaining the privacy implications of public hosting or third-party processing. Naming Makernice and requiring subjects' consent reduces concern, but consent to appearing in a video does not necessarily establish informed consent to public photo exposure and external processing. This is a privacy-disclosure gap rather than evidence of malicious exfiltration.

Content

No source excerpt is available for this finding.

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 38)May include surrounding context.

md
## 5. Generate after the user agrees

Only call `generate` after the user has seen the price and said yes. If the user already set a budget and told you to go ahead, you may start without asking again as long as `amount_usd` is within that budget; still use `amount_usd`, not the budget, as `max_cost_usd`.

- `max_cost_usd`: required. Use the approved `amount_usd` from `estimate_cost`, exactly as returned. The job is refused if the price is higher.
- `request_id`: required and unique for every new job, printable ASCII with no spaces, at most 64 characters, for example `trend-<template id>-1`. A used id can never be reused for a different job.