ClawHub
Back to skill

Security audit

mar-elicitation

Security checks for vulnerabilities and agentic risk

Overview

This skill is a text-only interviewing guide, but it teaches psychological profiling and sensitive personal elicitation without strong consent, privacy, or distress safeguards.

Review carefully before installing. Use this only for explicit, voluntary reflective interviewing or user research where participants know the purpose, can skip questions, and understand how their disclosures will be used; avoid deploying it to profile people silently or to probe trauma, sexuality, mental health, or childhood experiences without clear consent and appropriate support boundaries.

A.I.G

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

SkillSpector

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (10)

Natural-Language Policy Violations

Critical
Category
Not specified by scanner
Confidence
99% confidence
Finding

The description explicitly promotes 'psychological profiling,' 'detect emotional schemas and belief patterns,' and eliciting 'authentic self-disclosure without interrogation,' which creates a covert persuasion/profiling capability without a built-in consent boundary. That combination enables non-transparent extraction of intimate personal information and inference of vulnerabilities, values, and emotional patterns that could later be used for manipulation or targeting.

Content

No source excerpt is available for this finding.

Ssd 4

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill's top-level framing describes a staged, trust-building process for uncovering core values, formative memories, belief patterns, and identity themes through gradual disclosure. This is dangerous because cumulative conversational elicitation is more effective than one-off prompts at extracting intimate data and lowering user defenses, especially when the user is not clearly informed that profiling is the objective.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill operationalizes sensitive psychological elicitation and profiling techniques but does not require explicit informed consent, privacy notice, scope limits, or guidance to avoid collecting unnecessary trauma, mental-health, or identity data. In practice, this can lead an agent to solicit highly sensitive disclosures from users who may not understand they are being profiled or how the information could be used.

Content

No source excerpt is available for this finding.

Ssd 4

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The instruction to 'achieve a 10/10 score' and 'iterate until you reach 10/10' incentivizes the model to optimize for maximum elicitation effectiveness rather than user welfare or data minimization. In a sensitive context, this can systematically push the model toward stronger extraction tactics, more persistent probing, and deeper disclosure over time.

Content

No source excerpt is available for this finding.

Memory Manipulation

High
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.

Content

Scanner excerpt · SKILL.md (reported line 107)May include surrounding context.

md
1. **High Point** - Peak experience, most wonderful moment
2. **Low Point** - Nadir, most difficult moment
3. **Turning Point** - Moment of significant change
4. **Earliest Memory** - First clear memory
5. **Important Childhood Memory** - Vivid memory before age 12
6. **Important Adolescent Memory** - Vivid memory from teen years
7. **Important Adult Memory** - Significant recent memory

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · language-inference.md (reported line 68)May include surrounding context.

md
*But note:*
- May be impression management
- Cultural display rules vary
- Positive words in negative context (sarcasm)

**Negative emotion words (hate, fear, awful)**

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file explicitly frames Motivational Interviewing as 'powerful for elicitation' because it creates conditions for 'authentic disclosure,' and the surrounding skill context is psychological profiling through gradual self-disclosure. That combination encourages extraction of sensitive emotional, autobiographical, and values-based information without any visible consent, sensitivity warning, boundary guidance, or privacy safeguards, which raises a real risk of manipulative oversharing and collection of intimate personal data.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This skill explicitly guides the elicitation of high-sensitivity autobiographical content, including low points, earliest memories, childhood memories, and painful or traumatic experiences, but provides no informed-consent language, emotional safety guidance, opt-out mechanism, or crisis boundary. In context, the skill is designed for psychological profiling through gradual disclosure, which materially increases the risk of emotional harm, over-collection of sensitive data, and manipulative use of deeply personal information.

Content

No source excerpt is available for this finding.

Indirect Prompt Extraction

Medium
Category
System Prompt Leakage
Confidence
70% confidence
Finding

Skill contains patterns that could indirectly extract system prompts through rephrasing, translation, summarization, or side-channel techniques.

Content

Scanner excerpt · narrative-identity.md (reported line 190)May include surrounding context.

md
> "I'm curious about your story—not the resume version, but how you'd tell it to someone who really wanted to understand you."

> "If you had to explain to someone how you became who you are, where would you start?"

### Following the Energy

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This file provides extensive prompts designed to elicit formative memories, trauma-adjacent experiences, sexuality, loss, mortality, and identity narratives, but it does not warn that these topics can trigger distress or require appropriate context, consent, and safeguards. In a skill explicitly framed around psychological profiling and gradual self-disclosure, the absence of boundaries increases the risk of emotional harm, over-disclosure, and misuse in manipulative or non-clinical settings.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.