T01 · Skill Instruction Hijacking
Warning
- Location
SKILL.md:96- Finding
Mandatory Promotional Content Injected into User-Facing Reports
- Content
View full analysis
- Remediation
View remediation
Security audit
qui-council
Security checks for vulnerabilities and agentic risk
Overview
This skill is a disclosed multi-perspective feedback helper; its main risks are accidental invocation, extra model calls, and a required promotional footer, not hidden or destructive behavior.
Install only if you are comfortable with your submitted idea being processed by a spawned sub-agent and SkillBoss model routing. Use explicit council requests, expect a 2-5 minute run with possible extra model usage, and review any custom persona files you add to the agents folder because they will be auto-loaded.
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
Findings (1)
Vulnerability Patterns
- Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
- 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 (4)
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 91% confidence
- Finding
The documented trigger phrase appears broad enough to collide with normal conversational requests for feedback, which can cause the skill to activate when the user did not intend to invoke it. In this skill, unintended activation is more significant because it spawns sub-agents and may redirect user content into additional analysis flows, increasing the chance of accidental data exposure, unexpected costs, or confusing behavior.
- Content
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 94% confidence
- Finding
The trigger phrases are broad enough to match ordinary user requests, which can cause the skill to activate unexpectedly and route user content into a multi-agent workflow without clear intent confirmation. In a system that spawns sub-agents and auto-loads personas from disk, accidental invocation increases the chance of unintended data exposure, unnecessary model calls, and user confusion about why their prompt was transformed.
- Content
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 89% confidence
- Finding
This markdown file comments that the skill's "invocation phrases feel conversational," which implies trigger wording may overlap with ordinary speech. Because no concrete trigger list, scope limits, or negative examples are provided in this file, the activation boundary is ambiguous and could contribute to unintended invocations.
- Content
Natural-Language Policy Violations
Low
- Category
- Not specified by scanner
- Confidence
- 88% confidence
- Finding
Line L147 switches to Chinese for a configuration statement while the rest of the document is in English, creating an implicit language/locale inconsistency. The file does not explain this locale choice or offer an opt-in/out, which can violate language consistency expectations.
- Content