book-sft-pipeline

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.

Quality

72%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No known issues

Fix and improve this skill with Tessl

tessl review fix ./examples/book-sft-pipeline/SKILL.md

Quality

Content

77%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The content is highly actionable with a clear, validated multi-phase workflow, but it over-explains known concepts and keeps detail inline that would better serve progressive disclosure; a broken example reference also weakens navigation. Trimming rationale prose and moving large blocks to bundle files would improve both conciseness and structure.

Suggestions

Move the full SYSTEM_PROMPTS and PROMPT_TEMPLATES lists and the complete phase code into referenced files, keeping SKILL.md as a concise overview with pointers.

Remove or condense the Integration with Context Engineering Skills section and other rationale paragraphs that restate concepts Claude already knows.

Fix the broken ./examples/gertrude-stein/ reference by either adding the example bundle or removing the link.

Dimension	Reasoning	Score
Conciseness	The body is efficient in places with executable code and specific config, but explains concepts Claude already knows (base models are more malleable, OCR errors become learned patterns) and the Integration with Context Engineering Skills section adds length without operational value.	2 / 3
Actionability	Provides concrete copy-paste-ready code for each phase — extract_epub, segment, the Tinker training loop with real API calls, validation grep — plus specific config values and prompt templates.	3 / 3
Workflow Clarity	The six phases are explicitly sequenced with an architecture diagram, and Phase 6 supplies validation checkpoints (modern-scenario test, originality grep, AI detector) plus a Known Issues feedback section.	3 / 3
Progressive Disclosure	References are real and one-level deep (segmentation-strategies.md, tinker-format.md, tinker.txt, pipeline_example.py), but substantial content that could live in references (full template lists, full code, the cross-skill integration section) is inline, and the cited ./examples/gertrude-stein/ path does not exist.	2 / 3
	Total	10 / 12 Passed

Description

67%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is specific and occupies a distinct niche, but it relies on technical jargon for triggers and only implies activation conditions rather than stating concrete user-facing triggers. Rewording around natural user phrasings would raise trigger quality and completeness.

Suggestions

Add an explicit Use-when clause naming natural user triggers, e.g. when the user wants to fine-tune a model to write in an author voice, build a style-transfer dataset from a book, or train a LoRA on literary text.

Replace jargon-only terms (book-to-SFT, LoRA workflows) with phrasings a user would actually say, such as fine-tune a model on a book writing style and author-voice training.

Lead with the clearest user-facing framing of the task before listing the internal pipeline stages.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — ePub extraction, literary segmentation, author-voice dataset construction, style-transfer training, LoRA workflows, and model evaluation — matching the multi-action anchor.	3 / 3
Completeness	The what is explicit and detailed, but the when is only an implied capability list (should be used for book-to-SFT pipelines) rather than concrete user triggers, falling short of the explicit-trigger anchor.	2 / 3
Trigger Term Quality	Contains some natural terms (ePub, book, model evaluation) but leans on domain jargon (book-to-SFT, LoRA workflows, voice replication) that a user would rarely say verbatim, missing common phrasings like fine-tuning on a book style.	2 / 3
Distinctiveness Conflict Risk	The niche of book-to-voice-replication SFT is specific and unlikely to overlap with or trigger for unrelated skills.	3 / 3
	Total	10 / 12 Passed

Validation

87%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
relative_links	Relative link issues: 1 missing	Warning

	Total	14 / 16 Passed

Repository: muratcankoylan/Agent-Skills-for-Context-Engineering
Commit: c578e85

Reviewed: 7 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.