Security audit
Security checks for vulnerabilities and agentic risk
The skill is a coherent retail training quiz, but it stores identifiable employee performance records and shares them with managers without clear consent, access control, retention, or deletion rules.
Review this skill before installing in a workplace. It may be appropriate for staff training, but administrators should add clear notice to employees, manager-access rules, retention and deletion policy, and controls around when progress is written to memory.
The skill promises interactive training, employee-specific score tracking, and manager reporting, but the described implementation appears to only reference offline question generation and lacks real session/state controls. This mismatch can cause unsafe operator assumptions about what data is collected, stored, or shared, leading to privacy, compliance, and integrity issues when deployed.
The trigger list includes broad phrases like '测试', '练习', 'training', and '我想学', which are common in ordinary conversation and can cause accidental invocation. Unintended activation is risky here because the skill may start tracking employee progress or steer a conversation into training workflows without clear user intent.
The skill references a script and describes persistence of training records, yet it declares no explicit tool scope or permissions. That creates an overbroad/implicit capability boundary where file read/write behavior may be available without being transparently constrained, increasing the risk of unintended access to local data or unauthorized record storage.
The activation examples and interaction templates are predominantly in Chinese, and the session output script is explicitly written in Chinese without stating that the skill is China-specific or providing a language choice. This can violate language/locale policy when a skill implicitly enforces one language for all users.
The skill explicitly stores per-employee identifiers, session history, badges, and reports progress to managers, but it provides no warning, consent flow, or privacy notice. This creates a real privacy and employment-data handling risk, especially if staff are unaware that identifiable performance records are retained and shared.
This markdown file includes question templates and examples entirely in Chinese, such as the user-facing prompts on L010 and the examples on L015-L017, without stating that language selection is optional. Under the policy rule, forcing a specific language without user opt-in is a natural-language policy violation.
The generated quiz content is hard-coded in Chinese across multiple question templates and answer strings, but the script does not provide any language selection or explain that it is intentionally limited to a Chinese-only retail context. This creates a natural-language policy concern because it imposes a specific language on all users without opt-in.
Dynamic import() can load arbitrary modules at runtime, bypassing static analysis and potentially importing malicious code.
output_json = json.dumps(result, ensure_ascii=False, indent=2)
if args.output:
Path(args.output).write_text(output_json, encoding="utf-8")
print(f"✅ Generated {len(questions)} questions → {args.output}", file=__import__("sys").stderr)
else:
print(output_json)
No suspicious patterns detected.