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

55%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

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

Quality

Content

53%Weight 40%Scale 1-5

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

The content is a thorough, actionable pipeline with concrete code and clear file-based references, but it is verbose in places that restate known concepts and lacks explicit validation feedback loops for its batch/destructive phases. Progressive disclosure is well-organized aside from a missing referenced example directory.

Suggestions

Trim redundant concept explanations (e.g. 'Why Base Model?' rationale, the cross-skill 'Integration with Context Engineering Skills' section) and remove the duplicated rules between 'Critical Rules' and 'Guidelines' to respect the token budget.

Add explicit validation checkpoints and feedback loops (validate -> fix -> retry) after dataset construction and before/after training, matching the existing Phase 6 validation pattern.

Fix or remove the dangling ./examples/gertrude-stein/ reference and ensure all referenced bundle paths resolve, or move inlined API reference bulk into the existing tinker.txt reference.

Dimension	Reasoning	Score
Conciseness	The body is noticeably verbose with multiple padded sections restating concepts Claude already knows (e.g. 'Why Base Model?', the extended 'Integration with Context Engineering Skills' cross-references) and redundant 'Critical Rules'/'Guidelines' repetition.	2 / 5
Actionability	Provides mostly executable Python code across extraction, segmentation, prompt generation, dataset construction, and the Tinker training loop, with only minor gaps (e.g. llm_call and batch helpers undefined).	4 / 5
Workflow Clarity	The six phases are clearly sequenced and the validation phase exists, but batch/destructive operations like dataset construction and training lack explicit validation checkpoints and feedback loops, capping the score at 3.	3 / 5
Progressive Disclosure	Good structure with one-level-deep references to real bundle files (segmentation-strategies.md, tinker-format.md, tinker.txt) and a script (pipeline_example.py), though one referenced path (./examples/gertrude-stein/) does not exist and some API bulk is inlined.	4 / 5
	Total	13 / 20 Passed

Description

57%Weight 40%Scale 1-5

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 enumerates concrete capabilities clearly and occupies a distinct niche, but it omits an explicit 'Use when...' trigger clause, which caps completeness and limits trigger guidance. Trigger-term coverage is adequate but lacks natural synonyms and file extensions.

Suggestions

Add an explicit trigger clause such as 'Use when converting books or literary works into SFT fine-tuning datasets for style-transfer or author-voice models.'

Include natural user-facing terms and synonyms such as 'fine-tuning', 'ePub files', '.epub', and 'author voice' to improve trigger coverage.

Tighten the distinctiveness by naming the specific output (e.g. 'Tinker-compatible JSONL') to reduce overlap with generic SFT skills.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions (ePub extraction, literary segmentation, dataset construction, style-transfer training, LoRA workflows, evaluation), with only minor gaps in coverage relative to the full pipeline.	4 / 5
Completeness	Has a clear 'what' but no 'when' / 'Use when...' clause; trigger guidance is absent, capping completeness at 3 per the guideline.	3 / 5
Trigger Term Quality	Includes some relevant terms (ePub, segmentation, LoRA, SFT) but lacks the natural variations or file extensions a user might say, and omits common synonyms like 'fine-tuning'.	3 / 5
Distinctiveness Conflict Risk	The book-to-SFT / author-voice niche is fairly distinct with limited overlap risk against generic SFT skills, though 'SFT pipelines' is somewhat broad.	4 / 5
	Total	14 / 20 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
Path: examples/book-sft-pipeline/SKILL.md
Commit: 6dbe1a1

Reviewed: 19 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.