analyze-reference
拆解参考作品、提炼节奏、情绪、场景结构与禁用模式。Use when benchmarking imported novels from projects/<project-id>/references/raw/ and writing v2 outputs into source_assets/reference_analysis/.
Source facts
- Repository
- luoluomashang/novel-engine
- Last source activity
- April 27, 2026 at 02:35
- Detected SKILL.md language
- Chinese
- Stars
- 0
- Forks
- 0
Install options
The review-first prompt is selected by default. You can switch to a direct command or download a local copy.
Review the source files
Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.
Showing SKILL.md
SKILL.md
Source instructions · Read-only preview- name
- analyze-reference
- description
- 拆解参考作品、提炼节奏、情绪、场景结构与禁用模式。Use when benchmarking imported novels from projects/<project-id>/references/raw/ and writing v2 outputs into source_assets/reference_analysis/.
- argument-hint
- <project-id>
- disable-model-invocation
- true
# 拆解参考书
## 参数
- `$0`：`project_id`
## 执行步骤
1. 若 `$0` 缺失，立即停止并要求用户提供 `project_id`。
2. 先运行 `python app.py ensure-prereq "$0" reference_analysis`。
3. 调用 `style-miner` 子代理执行真实对标分析。
4. `style-miner` 必须直接写入：
- `projects/$0/source_assets/reference_analysis/reference_manifest.json`
- `projects/$0/source_assets/reference_analysis/reference_summary.json`
- `projects/$0/source_assets/reference_analysis/reports/<reference>.json`
5. 长文本至少覆盖前段 / 中段 / 后段；先剔除站点声明、上传说明、版权前缀。
6. 不要运行 `python app.py analyze-reference "$0"` 或 `scripts/analyze_reference.py` 来生成内容；这些脚本只保留给测试与回归。
7. 返回中文摘要，并建议进入 `/build-dna`。
## 输出要求
- `source_assets/reference_analysis/reports/*.json` 中的文本说明统一中文。
- 结论必须能支撑后续 DNA 编译，而不是只给宽泛评价。
## 阶段产物契约
- 输入文件：`projects/$0/references/raw/*`、已有的 `projects/$0/reference_info.json`。
- 输出文件：`projects/$0/source_assets/reference_analysis/reference_manifest.json`、`projects/$0/source_assets/reference_analysis/reference_summary.json`、`projects/$0/source_assets/reference_analysis/reports/*.json`。
- 禁止行为：不要运行 `python app.py analyze-reference` 或 `scripts/analyze_reference.py` 代写内容；不要把产物写回旧版 `analysis/` 或 `references/parsed/` 目录；不要只读开头几段就下结论。
View on GitHub