ProClaw Partner insight(优质伴侣分析技能)

• Outcome: Review
• Latest audit: May 29, 2026 at 9:35 PM UTC
• Version:
  1.0.0

Overview

This skill is technically simple, but it needs review because it encourages manipulative relationship tests, private partner tracking, and high-stakes relationship decisions from a rigid scoring framework.

Install only if you will treat it as a subjective journaling aid, not as a basis for secret testing or major relationship decisions. Avoid manufacturing crises, withholding affection or money to provoke reactions, or keeping detailed identifiable records without considering privacy and safety. For abuse, coercion, self-harm threats, legal issues, or urgent safety concerns, use qualified human support rather than this skill.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (17)

Context-Inappropriate Capability

Medium

• Confidence: 90% confidence
• Finding: The template explicitly encourages users to run 'boundary tests' such as overtime, withdrawal of investment, weakness exposure, and risk-decision tests on a partner. That moves the skill from passive reflection into active interpersonal probing/manipulation, which can escalate conflict, normalize covert testing, and cause psychological harm or retaliatory behavior in a sensitive relationship context.

Context-Inappropriate Capability

Medium

• Confidence: 93% confidence
• Finding: The file explicitly lists '在多个伴侣中选择' as an intended use case, which goes beyond analyzing a single relationship pattern and steers the skill into comparative romantic decision-making. In this context, that expansion increases the chance of manipulative, high-stakes advice affecting intimate relationships without adequate safeguards, consent considerations, or uncertainty handling.

Context-Inappropriate Capability

Medium

• Confidence: 96% confidence
• Finding: The guidance explicitly tells users to engineer interpersonal stressors ('pressure tests,' 'small crises,' and reduced material support) to observe a partner's reactions. In a relationship-advice skill, this moves from analysis into covert behavioral manipulation and deception, which can provoke emotional harm, escalate conflict, and normalize abusive testing behaviors.

Context-Inappropriate Capability

Medium

• Confidence: 95% confidence
• Finding: The examples operationalize manipulative tactics by recommending reduced meetings and material investment specifically to trigger reactions for evaluation. This makes harmful conduct easier to imitate and reinforces deceptive relationship testing as a normal diagnostic tool.

Vague Triggers

Medium

• Confidence: 87% confidence
• Finding: The trigger scope is broad enough to activate on generic relationship questions, which can cause the agent to deliver this skill's strong classification framework in contexts where the user did not explicitly ask for that model. In a relationship-advice skill that makes evaluative judgments, over-triggering increases the chance of inappropriate, overconfident, or harmful guidance being surfaced.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The skill provides structured judgments about a partner and recommends consequential actions such as whether to continue, improve, or 'stop loss' a relationship, but it does not adequately warn about misclassification, one-sided evidence, or the possibility of emotional harm from acting on the output. In this context, the framework can amplify bias and lead users to make major personal decisions based on reductive categories and incomplete information.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The FAQ explicitly tells users they do not need to tell a partner they are analyzing her and frames covert observation/recording as private decision support. In a relationship-analysis skill that already encourages categorizing and scoring partners, this increases the risk of deceptive surveillance, privacy invasion, and escalation of mistrust without any consent, safety, or ethical guardrails.

Missing User Warnings

High

• Confidence: 98% confidence
• Finding: The skill recommends 'creating pressure tests' by intentionally exposing weaknesses, manufacturing small crises, and pausing material support to observe the partner's reaction. These are manipulative provocation tactics that can emotionally destabilize the other person, trigger conflict, and normalize coercive relationship testing, especially dangerous in a skill focused on deciding whether a partner will 'drag down' or 'help' the user.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The template guides users through highly sensitive relationship scoring, partner classification, and breakup/continuation decision support without any warning about privacy, emotional harm, bias, or the limits of such heuristics. In this skill’s context, that increases the risk of overconfident, harmful personal decisions based on reductive labels and encourages users to record intimate data that could be mishandled or exposed.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: The document introduces 'pressure testing' as a recommendation without warning users about deception, emotional harm, or relationship damage. Omitting safety framing in a sensitive interpersonal context increases the chance that users will treat manipulative tactics as endorsed best practice.

Missing User Warnings

High

• Confidence: 98% confidence
• Finding: The instruction to deliberately expose weaknesses, create crises, and pause investment is explicit guidance for deceptive emotional provocation. In relationship advice, this can facilitate coercive dynamics, retaliation, psychological distress, and erosion of trust even when users follow the skill as intended.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: Repeating the harmful testing guidance in examples increases the likelihood of adoption because examples are often treated as actionable templates. The absence of caution compounds the risk by presenting emotional manipulation as practical, validated advice.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The document gives directive, high-stakes advice such as '立即止损' and '切断所有联系' in a quasi-diagnostic framework, but does not clearly warn that it is not a substitute for licensed legal, mental-health, or domestic-violence/safety guidance. In a relationship-assessment skill, users may over-rely on this framework for crisis decisions, which can lead to unsafe escalation, inappropriate isolation, or failure to seek qualified support when abuse, coercion, or mental-health emergencies are involved.

Natural-Language Policy Violations

Medium

• Confidence: 89% confidence
• Finding: The file consistently frames the risky partner as female ('她'), embedding a gender-specific assumption into risk detection and response. In this skill's context, that bias can distort assessments, alienate users, and reinforce stereotypes, causing the system to misclassify abuse dynamics or provide lower-quality guidance when the partner is male, non-binary, or when the user's situation does not match the assumed framing.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: This section explicitly recommends covert 'boundary tests' such as repeated late returns, pausing material support, and staged vulnerability/risk scenarios to probe a partner's reactions. In a relationship-advice skill, these are manipulative pressure tactics that can escalate conflict, normalize emotional testing, and create safety risks without any warning, consent requirement, or de-escalation guidance.

Ssd 4

Medium

• Confidence: 96% confidence
• Finding: The staged decision-tree format normalizes manipulative testing by embedding deception into a seemingly structured, authoritative workflow. That presentation lowers user resistance, turns unethical conduct into repeatable procedure, and can escalate controlling behavior in intimate relationships.

Ssd 4

Medium

• Confidence: 95% confidence
• Finding: The example section concretely demonstrates how to fabricate stress scenarios, making the manipulative pattern more actionable and easier to reproduce. In relationship contexts, such scripts can be used to justify emotional baiting and deceptive trust tests with real interpersonal harm.

VirusTotal

66/66 vendors flagged this skill as clean.