information-asymmetric-theory-of-mind-reasoning
Use this skill when the conversation involves multiple people and one person missed part of the discussion, so the agent has to reason about different perspectives instead of just the global truth. Trigger it for requests like 'who knows that?', 'what would she think happened?', or 'does he know the answer?'. Everyday examples include: someone leaving to get coffee and missing a key fact, a late joiner entering after the topic changed, and a character hearing only a partial recap instead of the original statement.
소스 정보
- 저장소
- Dingxingdi/paper_fast_search_backup
- 최근 소스 활동
- 2026년 4월 8일 15:14
- 감지된 SKILL.md 언어
- 영어
- 스타
- 0
- 포크
- 0
설치 방법
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
소스 파일 검토
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
파일 탐색기
4 개 파일SKILL.md 표시 중
SKILL.md
소스 지침 · 읽기 전용 미리보기- name
- information-asymmetric-theory-of-mind-reasoning
- description
- Use this skill when the conversation involves multiple people and one person missed part of the discussion, so the agent has to reason about different perspectives instead of just the global truth. Trigger it for requests like 'who knows that?', 'what would she think happened?', or 'does he know the answer?'. Everyday examples include: someone leaving to get coffee and missing a key fact, a late joiner entering after the topic changed, and a character hearing only a partial recap instead of the original statement.
# Skill: information-asymmetric-theory-of-mind-reasoning
## 1. Capability Definition & Real Case
* **Professional Definition**: The ability to reason about who knows what, and what a specific interlocutor would believe, when conversation participants have unequal access to information due to joining, leaving, or otherwise missing parts of a dialogue.
* **Dimension Hierarchy**: Social and Persona Modeling->Mental-state Reasoning->information-asymmetric-theory-of-mind-reasoning
### Real Case
**[Case 1]**
* **Initial Environment**: A group chat begins with Linda and David discussing Linda’s new dog. Kailey leaves to get coffee during the part where the breed is mentioned and returns only after the group shifts to talking about the dog’s tricks.
* **Real Question**: What breed would Kailey think Linda’s dog is?
* **Real Trajectory**: 1. Track which utterances occurred before Kailey left, during her absence, and after she returned. 2. Mark the dog-breed detail as inaccessible to Kailey. 3. Distinguish omniscient-view truth from Kailey-centric accessible knowledge. 4. Answer from Kailey’s perspective rather than from the full transcript.
* **Real Answer**: Kailey does not know the breed.
* **Why this demonstrates the capability**: The transcript contains the true breed, so a naive omniscient assistant can answer incorrectly by collapsing its own knowledge with the absent character’s mental state. Success requires modeling information accessibility and belief from a constrained perspective inside a live conversation.
## 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`
GitHub에서 보기