Back to skill

Security audit

OUA 统一智能评估

Security checks for vulnerabilities and agentic risk

Overview

The skill appears to be a legitimate AI assessment and report-generation tool, but its report generator can embed untrusted input into HTML without full escaping.

Install only if you are comfortable reviewing or patching the report generator. Do not generate or open HTML reports from result JSON supplied by untrusted parties until all report fields are safely escaped or a strict input schema is added.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/score_test.py:489
Finding

Stored HTML and Script Injection in Generated Reports

Content
View full analysis
📅 {session.test_date} 👤 {session.tester or 'Anonymous'} 🎯 {session.difficulty.upper()} 🔬 {session.track_mode.upper()} ⏱️ {session.duration_minutes} 分钟 ``` Several result fields are also rendered without escaping: ```python for result in session.results: info = DIMENSIONS[result.dimension] diff_map = {'basic': 'diff-basic', 'advanced': 'diff-advanced', 'expert': 'diff-expert'} diff_label = {'basic': '基础', 'advanced': '进阶', 'expert': '专家'} q_short = html.escape(result.question[:45]) + ('...' if len(result.question) > 45 else '') notes_html = html.escape(result.notes[:70]) if result.notes else '-' track_class = "track-oit-badge" if info["track"]=="OIT" else "track-lli-badge" track_label = "OIT" if info["track"]=="OIT" else "LLI" html_content += f""" {info['name_cn']} {track_label} {result.question_id} {diff_label.get(result.difficulty,result.difficulty)} {q_short} {result.score}/5 {notes_html} """ ``` These values can originate from an untrusted JSON document: ```python w ...[truncated 2737 chars]
Remediation
View remediation
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • 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 (5)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger phrase list is unusually broad, multilingual, and includes many generic benchmark/evaluation terms without clear invocation boundaries. This can cause accidental or overly aggressive skill activation in unrelated conversations, leading the agent to switch into this assessment workflow unexpectedly and potentially override the user’s intended task flow.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This markdown file presents all instructions, prompts, and scoring guidance in Chinese, effectively forcing a specific language for use of the skill content. Under the policy rule, a language-specific constraint should either be optional for the user or clearly documented as a justified region-specific limitation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

This code sets the generated report to lang="zh-CN" and uses a Chinese-only title and surrounding user-facing report text throughout the file. Under the policy, forcing a specific language/locale without user opt-in is a natural-language policy violation unless the constraint is explicitly justified.

Content

No source excerpt is available for this finding.

Tainted flow: 'output' from input (line 845, user input) → open (file write)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/score_test.py (reported line 848)May include surrounding context.

python
output = input("\n输出文件名 [默认: oua-report.html]: ").strip() or "oua-report.html"

    html_report = generate_html_report(session)
    with open(output, 'w', encoding='utf-8') as f:
        f.write(html_report)

    # 控制台总结

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
76% confidence
Finding

The main description and most user-facing instructional content are presented in Chinese, while English is only provided as a short alternate description field. This may effectively impose a language/locale preference without clearly offering the user a choice or stating that output can be provided in the user's preferred language.

Content

No source excerpt is available for this finding.