Model Context Transfer

模型切换时的上下文传递技能。当需要切换AI模型或Agent时，确保新模型能完整接收之前的对话记忆、任务状态和关键信息，实现无缝衔接。

Security Scan

The skill's files and instructions are consistent with its stated purpose (creating and exchanging structured context summaries when switching models/agents); it contains no unexplained installs, network endpoints, or credential requests.

Purpose & Capability

Name/description, SKILL.md, templates, and the small Python script all align: they generate and format a context-transfer document for handing work between models/agents. No unrelated binaries, env vars, or remote endpoints are requested.

Instruction Scope

Instructions are narrowly focused on producing and exchanging structured CONTEXT_TRANSFER documents. However the guidance explicitly mentions handing over '文件/路径/凭据' (files/paths/credentials) and recommends periodic automatic backups and triggers (e.g., '检测到上下文超过 80% 时'), which could encourage collection or transmission of sensitive data if operators do not redact it. The instructions do not themselves automate reading system files or secrets, but they give agents permission to include credential information in transfers if user/agent logic does so.

Install Mechanism

No install spec; skill is instruction-focused with one local Python script that generates markdown. No downloads, no external packages fetched by the skill itself.

Credentials

Skill declares no required environment variables or credentials and the script doesn't access environment secrets. Still, the quality checklist and templates mention transferring '凭据' (credentials) and file paths — this is a usage-level risk (user/agents could include secrets in the transfer) rather than a property of the skill's code.

Persistence & Privilege

always is false and skill does not request persistent elevated privileges. The included script writes a local markdown file (context_transfer.md), which is expected behavior for a document-generator; it does not modify other skills or system-wide settings.

Assessment

This skill appears internally coherent and does what it says: produce structured handover documents for switching models or agents. Before installing or using it, consider these precautions: - Secrets handling: the templates and checklist explicitly reference handing over credentials and file paths. Ensure you or your agent redact or exclude secrets (API keys, passwords, private file contents) from any CONTEXT_TRANSFER before sending it to another agent or model. - Triggering policy: the SKILL.md suggests automatic triggers (e.g., context thresholds, periodic backups). If you enable automation, limit when and to whom transfers are sent to avoid accidental data leakage. - Review the script: the Python script only formats and writes a markdown file locally. You can safely inspect it; run it in a controlled environment if you plan to execute it on sensitive data. - Agent permissions: restrict which agents/models can receive CONTEXT_TRANSFER documents and monitor any agents that are permitted to forward or export them externally. If you want higher assurance, ask the author to explicitly state policies for redaction, or modify the templates to include mandatory redaction placeholders for secrets and PII.

Model Context Transfer（模型切换上下文传递）

触发场景

• 切换不同AI模型（如 MiniMax-M2.7 → GLM-5）
• Agent A 交接任务给 Agent B
• QQ/微信等频道切换 Agent
• 长时间对话后模型记忆丢失

核心原则

1. 交接信息完整性

确保新模型接收：

• 项目背景：在做什么？
• 当前状态：做到哪了？
• 关键约束：不能做什么？优先级？
• 待处理事项：接下来要做什么？

2. 交接格式

用结构化的 CONTEXT_TRANSFER 格式传递：

## CONTEXT_TRANSFER

### 项目/任务
[简要描述]

### 当前状态
- ✅ 已完成：[列出]
- 🔄 进行中：[列出]
- ⏳ 待处理：[列出]

### 关键信息
- 关键决策：[记录]
- 约束条件：[记录]
- API/文件路径：[记录]

### 团队架构
[当前涉及的Agent及其职责]

### 下一步
[具体行动项，优先级排序]

3. 传递时机

• 主动触发：当用户说"切换模型"、"换Agent"、"让XX来处理"
• 被动触发：检测到上下文超过 80% 时，需要换模型续接
• 定期总结：每完成一个里程碑，主动输出 CONTEXT_TRANSFER 备份

使用方法

场景1：临时切换模型完成任务

当需要用不同模型处理时：

1. 当前模型生成 CONTEXT_TRANSFER 总结
2. 将总结发送给新模型
3. 新模型基于总结继续工作
4. 完成后反向传递结果

场景2：Agent 交接

当 Agent A 要把任务交给 Agent B：

1. Agent A 输出完整 CONTEXT_TRANSFER
2. Agent B 读取并确认理解
3. Agent B 继续执行

场景3：长期项目记忆维护

对于持续项目，定期更新：

• 每次重要决策后更新 MEMORY.md
• 使用 CONTEXT_TRANSFER 格式记录当前进度
• 确保任何模型接手都能快速上手

模板文件

快速交接模板

## CONTEXT_TRANSFER

**项目**: [项目名称]
**时间**: [交接时间]
**交接人**: [Agent名称]
**接收人**: [Agent名称]

### 📍 当前状态
- 进度：[百分比/阶段]
- 已完成：[]
- 进行中：[]
- 遇阻：[]

### 📌 关键信息
- 核心目标：[]
- 重要决策：[]
- 约束条件：[]
- 文件路径：[]
- API/工具：[]

### 👥 团队角色
- [Agent名]：负责XXX
- [Agent名]：负责XXX

### 🎯 下一步（优先级排序）
1. [优先级1]
2. [优先级2]
3. [优先级3]

### ⚠️ 注意事项
- 不能做：[]
- 敏感信息：[]
- 依赖项：[]

与 MEMORY.md 的关系

• MEMORY.md：长期记忆，项目全貌
• CONTEXT_TRANSFER：短期交接，当前状态

两者配合使用，确保模型切换无缝衔接。

质量检查清单

交接前确认：

• 项目背景清晰
• 当前进度明确
• 关键约束已记录
• 文件/路径/凭据已交接
• 下一步行动具体可执行
• 潜在风险已标注