1# Book SFT Pipeline
2 
3A standalone skill for training language models to write in any author's style. This is a **separate plugin** from the main Context Engineering collection.
4 
5## Installation
6 
7### Claude Code
8 
9```bash
10# Add the marketplace first
11/plugin marketplace add muratcankoylan/Agent-Skills-for-Context-Engineering
12 
13# Install the book-sft-pipeline plugin
14/plugin install book-sft-pipeline@context-engineering-marketplace
15```
16 
17### Cursor / Codex / IDE
18 
19Copy `SKILL.md` to your `.rules` or project skills folder.
20 
21### Manual
22 
23Reference the `SKILL.md` file directly in your agent's context.
24 
25## What's Included
26 
27```
28book-sft-pipeline/
29├── README.md                 # This file
30├── SKILL.md                  # Complete skill documentation (standalone)
31├── examples/
32│   └── gertrude-stein/       # Complete case study with real outputs
33│       ├── README.md         # Results and analysis
34│       ├── sample_outputs.md # Raw model outputs
35│       ├── training_config.json
36│       ├── dataset_sample.jsonl
37│       └── pangram/          # AI detector screenshots
38├── scripts/
39│   └── pipeline_example.py   # Conceptual implementation
40└── references/
41    ├── segmentation-strategies.md
42    ├── tinker-format.md
43    └── tinker.txt
44```
45 
46## Key Results
47 
48Trained Qwen3-8B-Base on Gertrude Stein's "Three Lives" (1909):
49 
50| Metric | Value |
51|--------|-------|
52| Training examples | 592 |
53| Loss reduction | 97% |
54| Pangram AI detector | 70% Human |
55| Training time | 15 minutes |
56| Total cost | $2 |
57 
58## Related Context Engineering Skills
59 
60This skill applies patterns from the [Agent Skills for Context Engineering](../../README.md) collection:
61 
62| Skill | Application |
63|-------|-------------|
64| [project-development](../../skills/project-development/) | Staged pipeline architecture |
65| [context-compression](../../skills/context-compression/) | Segmentation strategy |
66| [multi-agent-patterns](../../skills/multi-agent-patterns/) | Orchestrator pattern |
67| [evaluation](../../skills/evaluation/) | Modern scenario testing |
68| [context-fundamentals](../../skills/context-fundamentals/) | Prompt diversity |
69 
70## Resources
71 
72- [Dataset on Hugging Face](https://huggingface.co/datasets/MuratcanKoylan/gertrude-stein-style-sft)
73- [Research Paper](https://arxiv.org/pdf/2510.13939) (Chakrabarty et al. 2025)
74 
75## License
76 
77MIT
78 
79