← Back to Skills Marketplace
321
Downloads
0
Stars
0
Active Installs
17
Versions
Install in OpenClaw
/install zaomeng-skill
Description
面向中文小说人物蒸馏、关系抽取、关系图谱与角色对话的 ClawHub skill。
Usage Guidance
This package is internally coherent and appears to do what it claims: it prepares excerpts and prompt payloads, asks your host LLM to generate results, and then materializes persona files and exports relation graphs. Before installing: (1) verify you trust the skill source and run it in a controlled environment; (2) ensure Python and the optional dependencies (ebooklib, tiktoken) are installed from official sources; (3) only pass files you want read/stored — helper scripts read arbitrary paths and will write generated markdown/HTML into character folders and may append to MEMORY files; (4) inspect generated HTML/graph output before opening if you have concerns about embedded content; (5) do not provide secrets or credentials to this skill (it does not need them). If you want extra assurance, review tools/export_relation_graph.py and the full persona materializer code to confirm there are no remote network fetches or unexpected file-system writes in your specific bundle copy.
Capability Analysis
Type: OpenClaw Skill
Name: zaomeng-skill
Version: 4.1.3
The zaomeng-skill bundle is a comprehensive toolset for character distillation and roleplay from Chinese novels. It includes Python utilities for processing EPUB/TXT files, managing structured persona bundles in Markdown, and generating interactive relationship graphs. While `tools/_skill_support/relation_graph_export.py` utilizes `subprocess.run` to invoke a local headless browser (Edge/Chrome) for Mermaid-to-SVG rendering, the implementation avoids shell execution and uses sanitized inputs. The skill includes explicit safety policies in `references/safety_policy.md` and instructions in `SKILL.md` that are strictly aligned with its stated purpose, with no evidence of data exfiltration, credential theft, or malicious intent.
Capability Assessment
Purpose & Capability
Name/description describe character distillation and relation extraction; the bundle contains prompt templates, references, examples, and helper Python scripts (prepare excerpt, build payload, materialize persona, export graph) which are all appropriate and expected for this purpose.
Instruction Scope
SKILL.md instructs the host to prepare excerpts and run the included helper scripts which read user-supplied novel files and write generated persona/graph files. This scope is consistent with the stated purpose. Note: the helper scripts read arbitrary local file paths and write outputs (including appending to MEMORY files), so only supply texts you intend to be processed and stored.
Install Mechanism
No install spec is provided (instruction-only plus local helper scripts). The bundle includes a requirements.txt listing PyYAML and optional ebooklib/tiktoken — dependency installation is standard and proportionate.
Credentials
The skill declares no required environment variables, no credentials, and no special config paths. The runtime behavior (reading novel files, producing markdown/HTML outputs) matches that claim.
Persistence & Privilege
Flags show always:false and default autonomous invocation allowed (normal). The skill writes generated persona files into local directories and has a 'runtime_append' policy for MEMORY files — this is expected for a persona/memory workflow but worth noting because user corrections may be persisted locally.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install zaomeng-skill - After installation, invoke the skill by name or use
/zaomeng-skill - Provide required inputs per the skill's parameter spec and get structured output
Version History
v4.1.3
zaomeng-skill v4.1.3
- 新增 tools/materialize_persona_bundle.py 脚本，实现人物档案自动物化
- 新增 tools/_skill_support/persona_bundle.py 支持人物包处理
- SKILL.md 文档全面更新，详述产物、流程与宿主执行规范
- 工具链流程新增人物档案后处理阶段
- 明确区分 skill 与宿主职责，总结关键 helper 命令
v4.1.1
- 新增人物关系图谱导出工具（tools/export_relation_graph.py）。
- 增加 Mermaid 与 HTML 格式关系图谱的生成流程说明。
- 新增蒸馏及图谱生成进度播报规范，要求阶段性反馈，不输出冗余内部过程。
- 扩展 helper 脚本和命令说明，涵盖关系图谱导出。
- 明确任务完成后应给出可操作提示，如查看档案、关系图谱或进入对话模式。
v4.1.0
zaomeng-skill 4.1.0
- No code or asset changes detected in this release.
- Documentation reviewed; content unchanged.
v4.0.1
zaomeng-skill 4.0.1
- No file changes detected in this version.
- Documentation and process descriptions remain unchanged.
- No new features, fixes, or updates included.
v4.0.0
**Major refactor: Migrated zaomeng-skill to a lighter, LLM-host-first workflow with greatly simplified assets and utilities.**
- Reduced codebase: Removed nearly all runtime code and internal rule engines; only essential tools remain.
- Now focuses on preparing novel excerpts and assembling prompt payloads; all generation is delegated to the host LLM.
- Added two utility scripts: `prepare_novel_excerpt.py` and `build_prompt_payload.py` under the `tools` directory.
- Documentation and instructions streamlined to reflect LLM-first architecture and clarify agent invocation.
- No longer uses internal rules, templates, or local runtime modules—prompt organization and data prep only.
v3.3.0
**Major update: Transitioned to an LLM-first workflow with stricter model requirements and updated user guidance.**
- Now requires a usable LLM before running any distillation or chat process; rule-template outputs are no longer a fallback.
- Skill will not prompt users to configure `runtime/config.yaml` when host/agent already provides an LLM—redundant setup steps removed.
- New user-facing responses emphasize LLM availability before workflow begins and clearly explain halts if no model is found.
- Added internal support files: `relation_store.py`, `relation_visualization_exporter.py`, and `session_store.py`.
- Stronger focus on integrating with host LLM environments (OpenClaw, Hermes, etc.) instead of local/CLI-only configs.
- SKILL.md guidance, command patterns, and error handling fully revised for LLM-first best practices.
v3.2.0
- 引擎核心结构重构，新增 `runtime/src/core/cli_app.py`、`runtime/src/core/runtime_factory.py` 等核心模块，实现组件化和更清晰的运行时分层。
- 增加专用 logging 和异常处理支持：引入 `logging_setup.py`、`logging_utils.py`、`exceptions.py`，提升诊断与插件集成能力。
- 明确区分本地规则生成与可选 LLM 输出：配置灵活支持，用户可选择规则或自然台词输出，无需手动拼装角色链路。
- 运行时主入口与薄包装结构调整，相关文件说明及调用规范已同步更新。
- 文档：SKILL.md 全面同步新版结构和调用方式，补充复用说明与最佳实践。
v3.1.0
**zaomeng-skill v3.1.0 — 重大更新：内嵌运行时，弃用外部仓库依赖**
- 内置完整可运行 zaomeng 子集，无需再自动克隆外部 Git 仓库，运行时与数据均位于 skill 包内。
- 新增并整理多份 Markdown 文档与运行脚本（如 `INSTALL.md`, `MANIFEST.md`, `README.md` 等），实现更清晰的分层与依赖管理。
- 所有接口调用建议与命令行范例已统一迁移至 skill 包内的 `runtime/zaomeng_cli.py` 入口，废弃旧版 Git 仓库调用方式。
- 强化依赖说明，仅要求 Python 与 `PyYAML`（可选 `tiktoken`、`ebooklib`）；普通小说处理无硬性第三方依赖。
- 核心使用、标准流程与禁止行为规则已整体对齐新架构，禁止绕过 CLI 或
v2.1.1
zaomeng-skill 2.1.1
- 固定 zaomeng 仓库的源码版本，新增自动 checkout 到提交 649f7466738f99d60c454e167835462215cffc7d 的说明与命令
- 更新文档中的环境准备步骤，明确指定 clone 后需固定版本
- 其他使用说明、工作流、标准回复等未发生变化
v2.0.1
zaomeng-skill 2.0.1
- 增加了本地 zaomeng 引擎环境准备相关说明，明确自动克隆仓库和工作区判断流程。
- 明确禁止输出依赖排查、安装说明、源码路径等非用户视角的调试提示。
- 补充推荐面对用户的回复模板，覆盖正常流程和失败兜底情形。
- 优化工作流描述，要求仅在真实仓库准备后执行后续命令。
- 禁止将 prompt/schema/说明文件当作引擎替代，强化只用真实 CLI 的约束。
v2.0.0
zaomeng-skill 2.0.0 introduces a switch from JSON to Markdown for character profiles and relation data, with updated user instructions and workflow.
- Changed primary character and relation file format from JSON to Markdown (`.json` files replaced by `.md` files).
- Added guidance on new Markdown-based personality/character memory files in documentation.
- Updated chat usage instructions; clarified intent mapping for `act` and `observe` conversation modes.
- Streamlined rules on invoking and interacting with the skill via CLI.
- Expanded and reorganized explanation sections for natural language intent mapping and session management.
v1.0.9
zaomeng-skill 1.0.8
- Major documentation overhaul: SKILL.md rewritten for clarity and precision (Chinese, rules-focused).
- Clarifies that zaomeng-skill is a rules-driven, profile-based engine—not a generic chatbot.
- Adds strict CLI invocation requirements, with preferred patterns for chat and other commands.
- Explicitly lists prohibited behaviors and misuse of internal APIs/modules.
- Documents allowed commands for distillation, extraction, profile viewing, and correction memory.
- Tightens guidance for agent integration and operator confirmation.
v1.0.7
zaomeng-skill 1.0.7
- Clarified chat execution rules: one-shot chat commands are preferred by default, with interactive sessions only if explicitly requested.
- Added canonical one-shot command-line forms for chat operations.
- Updated guidance on offering starter turns when initiating interactive chat.
- No changes to functionality or extraction logic.
v1.0.6
zaomeng-skill 1.0.6
- Added interactive session rules for `/chat`, including mode selection (observe/act), character confirmation, and session initialization prompts.
- Clarified that `/chat` is treated as an interactive session, not a batch task.
- Introduced confirmation-gated execution for `/distill` and `/extract`.
- Updated instructions to explicitly guide operators through initial chat turns and artifact completion checks.
- No code or schema changes; documentation improved for interactive and tool-driven workflows.
v1.0.5
zaomeng-skill v1.0.5
- Added comprehensive usage examples, prompt templates, and validation/safety reference files for distillation, relation extraction, and correction workflows.
- Updated SKILL.md to specify a fully self-contained, phase-based workflow, including detailed input normalization, evidence extraction, profile and relation synthesis, and multilayer quality/safety gates.
- Documented explicit output contracts and trigger phrases/commands for invoking each feature.
- Introduced schema, safety policy, and validation guidelines to standardize outputs and enhance reliability.
- All required artifacts, prompt formats, and key references are now included in the skill package.
v1.0.1
- Clarified that the skill is a runtime adapter requiring the Dreamforge GitHub repository.
- Added detailed setup instructions, including new dependency and configuration steps.
- Added INSTALL.txt and MANIFEST.txt files for improved installation guidance.
- Updated SKILL.md with explicit usage and dependency notes.
v1.0.0
Initial release of zaomeng-skill:
- Distills character profiles and extracts relationship graphs from novel files (.txt/.epub)
- Supports immersive roleplay chat modes (observe/act) with OOC correction memory
- All outputs are stored locally as JSON—no cloud or OpenAI API required
- Provides core CLI commands for distill, extract, chat, view, and correction tasks
- Requires Python 3.10 or higher
Metadata
Frequently Asked Questions
What is Zaomeng Skill?
面向中文小说人物蒸馏、关系抽取、关系图谱与角色对话的 ClawHub skill。 It is an AI Agent Skill for Claude Code / OpenClaw, with 321 downloads so far.
How do I install Zaomeng Skill?
Run "/install zaomeng-skill" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Zaomeng Skill free?
Yes, Zaomeng Skill is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Zaomeng Skill support?
Zaomeng Skill is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Zaomeng Skill?
It is built and maintained by 王克斌 (@wkbin); the current version is v4.1.3.
More Skills