Back to skill

Security audit

subclip-project-orchestration

Security checks for vulnerabilities and agentic risk

Overview

The scanners report a benign verdict with high confidence: the reviewed skill contains one low-probability autonomy finding that was judged expected for its purpose, and no findings from the other scanner.

The scan supports a benign verdict with high confidence and reports complete coverage of the one Markdown component. Benign does not mean risk-free; consider whether the skill’s autonomous project-handling guidance fits your preferences before installing.

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
80% 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 17)May include surrounding context.

md
e used to guess an ID.
   - An `@Project Name` mention from an ordinary project chat carries the same exact `projectId`. Treat it as read-only cross-project context unless that request explicitly asks to edit it. Inspect only the tagged project's state, brief, chat, assets, captions, or frames relevant to the request; the currently open project remains the mutation target by default.
   - The host may also deterministically resolve a unique full project name, including an unambiguous minor typo, and include it in the `Tagged Subclip projects` block. That block is the exact-ID authority. When it is present, use its IDs immediately and never ask the user to tag the same project again.
   - A request may tag up to ten projects. Preserve the user's order when it conveys priority; otherwise process the exact set sequentially.
2. Inspection of another project is read-only and needs no grant.
3. A new `create_project` or `clone_project` result is automatically writable for this agent session.