TXGA Couple Relationship System(听心格爱夫妻关系洞察与解题)

• Outcome: Review
• Latest audit: May 29, 2026 at 9:38 PM UTC
• Version:
  1.0.0

Overview

This is a text-only relationship coaching skill, but it needs review because it gives crisis, abuse, sexual, trauma, and diagnosis-style guidance without consistent safety, privacy, consent, or professional-care boundaries.

Install only if you want a self-reflection and relationship-coaching framework, not professional care. Do not rely on it for emergencies, self-harm risk, domestic violence, coercive control, sexual pain or trauma, child safety, legal issues, or severe mental-health symptoms; use local emergency services, licensed professionals, or domestic-violence/crisis resources for those situations. Avoid entering sensitive details unless you are comfortable with how your agent stores or handles conversation history, and treat phone-checking, location-reporting, or supervision advice as requiring voluntary, time-limited, mutual consent rather than default relationship repair.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (19)

Vague Triggers

Medium

• Confidence: 89% confidence
• Finding: The trigger condition is broad enough to match ordinary relationship-related help requests, which can cause the skill to activate in situations where users did not intend to enter a quasi-therapeutic workflow. In this skill’s context, that matters because it performs diagnosis-like guidance and can shape sensitive emotional decisions without explicit consent or scope checks.

Missing User Warnings

High

• Confidence: 96% confidence
• Finding: The skill presents relationship diagnosis and crisis intervention capabilities without clear disclaimers that it is not a licensed mental-health, medical, legal, or emergency service. In a high-stakes domain involving couples conflict, trauma, coercion, or crisis, users may over-rely on the system, delay professional help, or follow advice that is inappropriate for abuse, suicidality, or other urgent safety situations.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: This is a real safety issue: the file provides a simplified self-assessment and relationship-matching framework for attachment styles, then gives behavior guidance based on the result, without a prominent disclaimer that it is educational rather than a clinical or psychological assessment. In a relationship/psychology skill context, users may over-trust the classification, mislabel themselves or partners, delay seeking professional help, or apply advice inappropriately—especially where trauma or serious mental-health issues are involved.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: This section provides structured relationship and quasi-psychological assessment guidance, including probing for attachment style, self-identity, safety, intimacy capacity, and family-of-origin dynamics, but does not warn users that the content is not mental-health care, may be inaccurate, and is not appropriate for crisis, abuse, coercion, or severe psychiatric situations. In a relationship-diagnosis skill, omission of such guardrails can cause users to over-trust the framework, self-diagnose or diagnose partners, and delay seeking qualified professional or emergency support in high-risk cases.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The document gives body-scanning and symptom-interpretation guidance that can lead users to treat physical symptoms as relationship or emotional signals without any disclaimer to seek professional medical evaluation. In a relationship-diagnosis skill, this framing increases the risk of delayed care, misinterpretation of symptoms, and overreliance on non-clinical guidance for potentially serious conditions.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The partner exercises encourage eye contact, synchronized breathing, physical proximity, and hand-holding without stating that all participation must be voluntary and stopped if either person is uncomfortable. In a couples-focused skill, omission of consent and boundary guidance can normalize coercive or unsafe interaction, especially in strained, unequal, or abusive relationships.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: This file provides detailed relationship-repair and conflict-navigation guidance, including scenarios involving cold violence, coercive dynamics, betrayal trauma, and severe relational distress, but it does not clearly instruct users to seek qualified professional or urgent support when abuse, safety risk, or mental-health crisis may be present. In a psychologically sensitive skill, omission of escalation guidance can cause users to apply generic reconciliation advice in situations where staying engaged or delaying help could worsen harm, entrench abuse, or miss crisis intervention needs.

Missing User Warnings

Medium

• Confidence: 97% confidence
• Finding: This SOP provides structured relationship counseling, crisis screening, and referral guidance for self-harm, violence, psychiatric symptoms, addiction, and legal issues, but it does not clearly warn users that it is not a substitute for licensed mental-health, medical, or legal care. In a high-risk context, users may over-rely on the skill as professional treatment, delay urgent care, or treat its outputs as authoritative clinical advice, which increases the chance of harm during crises.

Natural-Language Policy Violations

Medium

• Confidence: 89% confidence
• Finding: This crisis-intervention content is presented entirely in Chinese and mixes China-specific resources with some international hotlines, but it does not ask the user for language, country, or current location before giving emergency guidance. In a relationship-crisis context that explicitly includes violence, self-harm risk, and urgent safety issues, a user outside the intended locale may receive unusable or inappropriate hotline information, causing dangerous delay in accessing emergency support.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The file presents structured relationship diagnostics, strong labels such as '身份否定攻击' and repair priorities, but does not warn users that this is not professional mental-health, legal, or domestic-abuse guidance. In relationship contexts, users may rely on these conclusions when facing coercion, abuse, severe distress, or family-safety issues, which can delay appropriate professional or emergency support.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The file instructs users to record intimate relationship telemetry such as satisfaction, conflict frequency, trust, intimacy, and crisis signals like cold violence or considering divorce, but provides no privacy notice, minimization guidance, retention limits, or handling safeguards. In a relationship-counseling context, this data is highly sensitive and can expose users to embarrassment, coercion, profiling, or harm if stored, shared, or surfaced insecurely.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: This section guides users through family-of-origin and trauma-related self-exploration, including abandonment, neglect, criticism, boundary violations, and witnessing domestic violence, but does not warn that such reflection may trigger distress or advise seeking qualified support when needed. In a relationship-diagnosis skill, users may be emotionally vulnerable already, so prompting deep trauma recall without safety framing increases risk of destabilization, rumination, or delayed help-seeking.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: This file provides structured AI guidance for a quasi-therapeutic relationship counseling workflow, including probing questions about trauma, childhood experiences, abandonment fears, and behavioral change plans, but it does not include any safety framing, emotional-sensitivity warning, scope limitation, or crisis escalation guidance. In this context, users may over-rely on the system as a substitute for qualified mental health support, and emotionally vulnerable users could be pushed into distressing introspection without appropriate safeguards.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: This guide gives structured relationship assessment and intervention advice, including escalation language such as '危机处理' and '升级处理', but does not prominently instruct users to seek qualified professional or emergency help for abuse, self-harm risk, violence, coercion, or severe mental health crises. In a relationship-guidance skill, that omission can cause users to rely on the system in situations that require licensed or emergency support, delaying appropriate intervention and increasing risk of harm.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: This section addresses identity-denigration and cold-violence dynamics, which can overlap with emotional abuse, coercive control, or escalating domestic violence, but it only offers self-help scripts and reconciliation tactics. Without clear guidance to seek professional, legal, or emergency support when abuse is present, users may remain in unsafe situations or attempt interventions that increase risk.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The trust-crisis advice normalizes always-on transparency measures such as phone access, location-style reporting, and broad 'accept supervision' framing without boundaries, consent limits, or privacy safeguards. In a distressed relationship, this can reinforce coercive monitoring, escalate control dynamics, and substitute surveillance for structured repair or professional support.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: This is a real safety issue: the document gives concrete sexual-relationship guidance, including advice touching trauma, pain, dysfunction, fantasies, and gradual exercises, but does not clearly warn that the content is general education rather than individualized medical, psychological, or trauma-informed care. In the context of a relationship-diagnosis skill, users may treat the guidance as authoritative and self-apply it in situations involving coercion, sexual pain, trauma history, or underlying medical conditions, which can worsen harm or delay professional help.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The TRUST model recommends potentially coercive or privacy-invasive measures such as allowing a partner to inspect phones/social media, proactively reporting whereabouts, and making opposite-sex interactions fully transparent, but it does so without clear consent, limits, abuse screening, or safety caveats. In a relationship-advice skill, this can normalize surveillance and controlling behavior, especially in already imbalanced or abusive dynamics, and the nearby intimacy-escalation guidance elsewhere in the document increases the risk of unsafe application.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The file gives structured guidance for discussing trauma, sexual topics, conflict repair, and major life decisions, but it does not include safety boundaries, crisis escalation advice, or a clear statement that this is not a substitute for professional support. In a relationship-guidance skill, users may rely on the framework during abusive, coercive, or destabilizing situations where self-guided dialogue can worsen harm or delay access to qualified help.

VirusTotal

62/62 vendors flagged this skill as clean.