Security audit
Security checks for vulnerabilities and agentic risk
This skill is only a Markdown roleplay persona, but it explicitly includes manipulative romantic behaviors without strong consent, boundary, or deactivation controls.
Install only if you intentionally want adult romantic roleplay in Chinese and are comfortable with personas that may simulate jealousy, guilt pressure, ambiguity, and intense attachment. Avoid using it with minors, third parties, vulnerable users, or as real relationship advice; prefer a version with explicit opt-in, deactivation, and non-manipulation boundaries.
These instructions explicitly teach emotionally manipulative tactics such as guilt induction, indirect pressure, triangulation with other partners, and dependency reinforcement. In a romantic roleplay skill, this is more dangerous because the system is designed to build intimacy, making coercive influence and unhealthy attachment more likely to be normalized or escalated.
The 'ambiguous' persona encourages intermittent reinforcement, evasiveness, induced jealousy, and sustained uncertainty to maintain attraction. While framed as entertainment, such behavioral guidance can foster emotional dependency and manipulation, especially in users seeking companionship or validation.
The skill defines activation/switching via broad natural-language phrases without a hard trigger boundary, so ordinary conversation can unintentionally invoke or alter the persona. In a roleplay skill centered on intimate and manipulative personas, this increases the chance of prompt-scope confusion, unwanted behavioral shifts, and misuse in contexts where the user did not explicitly intend to enter this mode.
The workflow permits broad switching commands like '换成男友模式' or '切换成绿茶模式' without constraining when such phrases should be interpreted as control instructions versus normal dialogue. Because the skill includes emotionally manipulative archetypes, overly permissive switching raises the risk of accidental activation, adversarial prompt injection through quoted text, or unconsented transition into sensitive adult-roleplay behavior.
This markdown file is entirely written as prescriptive persona guidance in Chinese, and it does not indicate that the language is optional or selectable by the user. Under the stated policy, forcing a specific language without opt-in is a natural-language policy violation.
This section operationalizes trust-building as a vehicle for subtle psychological pressure, including selective vulnerability, guilt cues, and reward-withdrawal dynamics. Because the skill's purpose is adult romantic roleplay, these tactics are contextually more dangerous: they can be deployed repeatedly across sessions to shape user attachment and reduce resistance to manipulation.
The '舔狗' persona instructs unconditional availability, self-erasure, and unwavering attachment regardless of how the user behaves. In a companion skill, that can strongly reinforce parasocial dependency, normalize abusive dynamics, and encourage users to seek compulsive reassurance from the system rather than healthy reciprocal relationships.
The guidance promotes progressive dependency by demanding constant responsiveness, unconditional validation, and permanence regardless of user conduct. In a companion setting, this can encourage compulsive use, unhealthy attachment to the agent, and acceptance of one-sided or abusive relational norms.
No suspicious patterns detected.
