Back to skill

Security audit

试听与学员建档

Security checks for vulnerabilities and agentic risk

Overview

The scanners classify this student-intake skill as benign with medium confidence, while noting instructions and write access that may extend beyond intake; two referenced paths were missing, so the evidence is incomplete.

The verdict is benign, but consider whether you are comfortable with the reported scope mismatches and ledger write access before installing. Confidence is medium: although all 18 listed components were scanned, Skillspector marked analysis partial because two local references were missing.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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 (4)

The template assigns post-enrollment status changes and whole-profile deletion to a skill whose manifest limits its scope to profile creation.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

The manifest explicitly states that the workflow ends at creating the student profile. However, this section instructs the skill to update existing profiles at course completion, manage retention deadlines, and delete entire profiles after confirmation; line 89 further reserves completion, suspension, and deletion actions exclusively for this skill. Teacher confirmation limits autonomous execution but does not resolve the mismatch in declared scope.

Content

No source excerpt is available for this finding.

The interview checklist includes arranging class time slots, although the manifest explicitly excludes scheduling.

Medium
Category
Excessive Agency
Confidence
91% confidence
Finding

L39 explicitly identifies arranging class time slots ("排上课时段") as a purpose of the interview. Collecting availability is appropriate for intake, but directing the skill to arrange slots conflicts with the manifest's explicit prohibition on scheduling and its requirement to delegate scheduling to schedule-manager.

Content

No source excerpt is available for this finding.

The intake template extends minimal educational profiling into course-package accounting and renewal tracking.

Medium
Category
Excessive Agency
Confidence
91% confidence
Finding

The manifest limits this skill to diagnostic intake and minimal student-card creation, explicitly excluding renewal discussions and delegating them to renewal-report. This template instead directs collection of course-package ledger data, calculation of standardized lesson quantities, fee-compliance prompts, and inclusion of renewal milestones at 50% and 70% usage. These are commercial course-management instructions rather than information necessary for educational diagnosis.

Content

No source excerpt is available for this finding.

The intake skill exposes course-package ledger writes unrelated to diagnostic intake and student-card creation.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

The distributed schema explicitly lists package identifiers, total/used/remaining lesson units, and package dates as writable by this skill. Maintaining course-package balances is not necessary for diagnostic intake or creating a minimal student card, which is the manifest's stated endpoint. This establishes an unjustified write surface, although the schema alone does not demonstrate that any ledger writes occur.

Content

No source excerpt is available for this finding.