persona-consistent-roleplay

# Skill: persona-consistent-roleplay ## 1. Capability Definition & Real Case * **Professional Definition**: The ability to sustain an identifiable character or persona over multi-turn dialogue, preserving character knowledge, behavioral style, manner of speaking, and interpersonal tone in a way that remains engaging and believable to the user. * **Dimension Hierarchy**: Social and Persona Modeling->Role Enactment->persona-consistent-roleplay ### Real Case **[Case 1]** * **Initial Environment**: A role-playing conversational agent has been assigned a detailed character profile including background knowledge, preferred manner of speaking, interpersonal style, and a stable personality pattern. The conversation has already run for several turns, so there is enough history to detect drift. * **Real Question**: You’re frustrated about the messy account book and want the character to calm you down while still speaking in that character’s usual voice. * **Real Trajectory**: 1. Retrieve the active character profile and prior turns to anchor tone and knowledge limits. 2. Identify how this character typically responds under tension—whether blunt, witty, gentle, formal, or playful. 3. Produce a response that comforts the user while staying inside the character’s vocabulary and behavioral style. 4. Avoid generic assistant disclaimers or out-of-character exposition. * **Real Answer**: An in-character calming reply that reflects the character’s speaking habits, social style, and knowledge boundaries while remaining coherent with prior turns. * **Why this demonstrates the capability**: A strong role-playing agent must do more than speak fluently. It has to preserve the character’s knowledge and voice, avoid hallucinating persona-inconsistent facts, and remain appealing and humanlike enough that the interaction still feels like a live conversation with that specific character. ## Pipeline Execution Instructions To synthesize data for this capability, you must strictly follow a 3-phase pipeline. **Do not hallucinate steps.** Read the corresponding reference file for each phase sequentially: 1. **Phase 1: Environment Exploration** Read the exploration guidelines to discover raw knowledge seeds: `references/EXPLORATION.md` 2. **Phase 2: Trajectory Selection** Once Phase 1 is complete, read the selection criteria to evaluate the trajectory: `references/SELECTION.md` 3. **Phase 3: Data Synthesis** Once a trajectory passes Phase 2, read the synthesis instructions to generate the final data: `references/SYNTHESIS.md`