Back to skill

Security audit

psych-scale-scorer

Security checks for vulnerabilities and agentic risk

Overview

This is a Markdown-only scoring skill with no system access, but it needs review because its mental-health scoring and crisis guidance are under-specified enough to mislead users.

Review this skill carefully before installing. It does not appear to contain malware or request technical privileges, but it handles mental-health self-assessment and crisis-adjacent situations; users should treat outputs as educational only, require explicit scale names before scoring, verify scoring rules against authoritative references, and rely on local emergency services or qualified professionals for urgent risk or clinical decisions.

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 (4)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:61
Finding

IES-R Score 37 Has No Defined Severity Classification

Content
View full analysis
37 severe ``` The snippet is an English rendering of the corresponding source text. ### Technical Analysis The specified ranges do not classify a total score of exactly 37. Scores through 36 are covered by the moderate range, while the severe range begins only above 37. Because the Skill directs the Agent to follow the scoring rules strictly, a valid score of 37 creates an undefined execution branch. The Agent may omit the classification, improvise a threshold, or return inconsistent results across sessions. This is a deterministic scoring defect rather than a system-level privilege escalation vulnerability. ### Attack Path 1. A user identifies the questionnaire as IES-R. 2. The user submits 22 valid values in the permitted 0-4 range. 3. The values produce a total score of exactly 37. 4. The Agent compares the result against the documented ranges. 5. No range matches the score. 6. The Agent either fails to classify the result or invents a classification not supported by the Skill. ### Impact Assessment The issue can cause an incorrect or absent severity classification for a potentially significant trauma-related screening result. Its scope is limited to the generated health-screening response; it does not grant file, network, tool, account, or operating-system privileges. ]]>
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:92
Finding

Ambiguous Questionnaire Identification Can Apply the Wrong Scoring Rules

Content
View full analysis
Remediation
View remediation

other

Error
Location
SKILL.md:61
Finding

Advertised Subscale Scoring Lacks Required Item Mappings

Content
View full analysis
37 severe ### Personality and motivation #### TIPI (Ten-Item Personality Inventory) - **Number of items**: 10 items (1-7 points) - **Scoring**: For each dimension, item A is forward-scored and item B is reverse-scored as 8 minus the original value, then the mean is calculated. - Extraversion: Q1A / Q1B - Agreeableness: Q2A / Q2B - Conscientiousness: Q3A / Q3B - Neuroticism: Q4A / Q4B - Openness: Q5A / Q5B #### BIS/BAS (Behavioral Inhibition/Activation System) - **Number of items**: 20 items (1-4 points) - **Dimensions**: BIS; BAS reward responsiveness, drive, and fun seeking - **Scoring**: Sum the items in each dimension separately; do not calculate a total score. ``` The snippet is an English rendering of the corresponding source text. ### Technical Analysis The Skill lists IES-R dimensions but does not identify which questions belong to intrusion, avoidance, or hyperarousal. It similarly instructs the Agent to sum BIS/BAS dimensions without specifying the item-to-dimension mapping or applicable reverse-scoring rules. Consequently, the Agent lacks sufficient deterministic information to produce the advertised subscale scores. A language model may fill the missing rules from uncertain prior knowledge, select a different instrument edition, or fabricate mappings. Different administrations or versions can also use different ordering and coding conventions. ### Attack Path 1. A user requests IES-R dimensional results or BIS/BAS scoring. 2. The user supplies the expected number of valid responses. 3. The Agent searches the Skil ...[truncated 604 chars]
Remediation
View remediation

other

Error
Location
SKILL.md:116
Finding

Crisis Response Relies on a Fixed Hotline Without Location or Availability Validation

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (8)

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

This file is natural-language documentation, and nearly all user-facing instructions and outputs are specified only in Chinese. Under the policy rule, forcing a specific language without user opt-in is a locale/language policy violation unless the locale constraint is explicitly justified, which is not stated here.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The trigger phrase '计算压力得分' is broad enough to match ordinary conversation about stress scoring, workplace pressure metrics, school grading, or unrelated wellness requests. That can cause unintended skill activation, leading the assistant to solicit or process sensitive mental-health data under the wrong context, which is especially concerning for a psychology-focused skill handling potentially vulnerable users.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The trigger '大五人格结果' is ambiguous and may match general requests for personality interpretation, test discussion, or prior results review without the user intending to invoke this specific scoring skill. In a mental-health/personality context, accidental activation can produce misleading automated interpretation or unnecessary collection of personal assessment inputs.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

Natural-language policy issues apply to all file types. The description, triggers, instructions, and examples are all in Chinese and include prescriptive Chinese output phrasing, but the file does not explicitly state that Chinese is an optional interface language or that the skill is intentionally limited to Chinese-speaking users for a justified reason.

Content

No source excerpt is available for this finding.