T01 · Skill Instruction Hijacking
Warning
- Location
SKILL.md:24- Finding
Fetched Web Content Is Not Isolated Against Indirect Prompt Injection
- Content
View full analysis
- Remediation
View remediation
Security audit
ai-high-quality-briefing
Security checks for vulnerabilities and agentic risk
Overview
The skill is a coherent Chinese AI/tech briefing assistant, but it allows use of logged-in browser pages for research without clear consent or safeguards.
Review before installing. Use this skill for public AI/tech briefing work, and avoid letting it inspect logged-in, private, internal, or subscription pages unless you explicitly choose the exact page and are comfortable with its contents being summarized or reused. If installed, prefer public web_search/web_fetch sources and disable or tightly scope any recurring reminder.
Vulnerability Patterns
- Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
- Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
- 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 (2)
T05 · Unauthorized Access and Privilege Escalation
Warning
- Location
SKILL.md:27- Finding
Authenticated Browser Content May Be Accessed Without a Least-Privilege Boundary
- Content
View full analysis
- Remediation
View remediation
Vulnerability Patterns
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
- 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 (4)
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The skill description is extremely broad and matches many ordinary research, summary, newsletter, and briefing requests, which can cause the agent to activate this skill in situations where it was not specifically intended. Over-broad activation increases the chance of inappropriate tool use, scope capture, and instruction takeover from a generic content request into a web-research workflow.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
Mandating Chinese output without checking the user's preferred language can override user intent and create an instruction-priority conflict, causing the skill to produce responses in an unexpected language. This is primarily a safety and usability issue, but in an agent setting it can also interfere with downstream workflows that expect another language.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
The writing-style section repeats a hard requirement for Chinese output, reinforcing behavior that may conflict with user instructions and making the override more persistent across the skill. Repeated mandatory language constraints increase the likelihood that the agent ignores user preference or system-level localization expectations.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 88% confidence
- Finding
This markdown file mixes English headings with core scoring criteria written in Chinese, and later sections also prescribe Chinese-language output formats. That creates a language/locale constraint without any user opt-in or explanation that the skill is specifically intended for a Chinese-language audience.
- Content
Static analysis
No suspicious patterns detected.