book-sft-pipeline

An agent skill by muratcankoylan, from muratcankoylan/Agent-Skills-for-Context-Engineering. Tags: audio, documents, evaluation, generative, mlops.

What it does

This skill should be used for book-to-SFT pipelines: ePub extraction, literary segmentation, author-voice dataset construction, style-transfer training, LoRA workflows, and model evaluation for voice replication.

Install

With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:

npx skills add muratcankoylan/Agent-Skills-for-Context-Engineering --skill book-sft-pipeline

Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):

git clone --depth 1 https://github.com/muratcankoylan/Agent-Skills-for-Context-Engineering
cp -r Agent-Skills-for-Context-Engineering/examples/book-sft-pipeline ~/.claude/skills/book-sft-pipeline

Safety box score

Grade B, low risk: 23 of 100 risk points, rated 2026-10-19. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.

This skill builds book-to-SFT pipelines: it extracts ePub text, segments it, sends excerpts to external LLM/detector services, and trains LoRA adapters for author-voice replication. The main risks are outbound transfer of book text to third-party APIs, local code execution during preprocessing/training, and a few unreadable assets that limit review.

Category levels, from 0 (none) to 3 (high)
Category	Level	Why
Code & shell execution	1 (low)	It ships runnable-looking Python in `scripts/pipeline_example.py` and shell install commands in `README.md`, but there is no auto-executing installer or `curl \| sh`/`eval` flow.
Network access	2 (moderate)	The pipeline explicitly calls remote services for generation and validation, including `llm_call(...)`, `service_client.create_lora_training_client_async(...)`, and detector services like GPTZero/Pangram/ZeroGPT.
Filesystem reach	1 (low)	It reads local ePub files and writes datasets/training artifacts, and the install steps copy the skill into hidden config directories such as `.claude/skills`, `.cursor/skills`, and `.agents/skills`.
Secrets & credentials	0 (none)	There is no request to read or paste passwords, tokens, `.env` files, keychains, or browser/CLI credentials.
Privilege & persistence	0 (none)	No `sudo`, service registration, global package install, or OS/shell startup modification is shown.
Instruction hijack surface	1 (low)	The skill embeds persona/system prompts like `You are an expert creative writer capable of emulating specific literary styles` and author-voice templates; this is core to the task, but it does steer the agent’s role.
Irreversible actions	0 (none)	It does not publish, deploy, send messages, book, pay, or delete anything; it only prepares data, trains, and validates.
Transparency	1 (low)	Three PNG screenshots in `examples/gertrude-stein/pangram/` are binary, and `references/tinker.txt` was not read due size/budget, so part of the reference material is not directly reviewable.

Before installing

Read `SKILL.md` and `scripts/pipeline_example.py` to confirm the exact preprocessing/training flow you’d actually run.
If you will use external LLMs or detectors, verify the text you send is acceptable to share off-machine.
Test first on a public-domain ePub in a scratch project and inspect the generated JSONL before using any proprietary or sensitive books.
If you do not want hidden config changes, avoid the `.claude/.cursor/.agents` copy steps or do them in a sandbox.

Rated 2026-10-19 by an automated review of 8 files from the skill's repository.

Source

Repository: muratcankoylan/Agent-Skills-for-Context-Engineering (all skills from this repository)
Path: examples/book-sft-pipeline/SKILL.md
Branch: main
Updated: 2026-09-19

fine-tuning-expert — Use when fine-tuning LLMs, training custom models, or adapting foundation models for specific tasks.
add-read-aloud — Use when the user runs /add-read-aloud or wants the app to speak text with Grok text-to-speech: read-aloud button on assistant replies, auto-speak, TTS.
add-voice — Use when the user runs /add-voice, types Voice Mode, or asks to add Grok realtime voice to an app.
arena — Spawn N parallel candidates at the same task, pick a base, graft the strongest parts of the losers into it.
interpreting-culture-index — Interprets Culture Index (CI) surveys, behavioral profiles, and personality assessment data.
canvas-design — Create beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art.