Skip to content
Back to skills

Fine Tuning Workflows

ASecurity

Principal-level fine-tuning lifecycle — when fine-tuning beats prompting + RAG, dataset curation, instruction tuning vs preference optimisation (SFT / DPO / RLHF), parameter-efficient methods (LoRA / QLoRA / adapters), evaluation, safety re-tuning, deployment, monitoring, and the cost / benefit framework for choosing between fine-tuning, RAG, and base-model usage.

  • 12 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 6, 2026
ai-agentsgo

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned October 6, 2026

npx -y skills add Nmor/the-council --skill fine-tuning-workflows --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Fine Tuning Workflows?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Fine Tuning Workflows
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/nmor-fine-tuning-workflows-the-council/badge)](https://www.skillsdirectory.com/skills/nmor-fine-tuning-workflows-the-council)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: fine-tuning-workflows
description: Principal-level fine-tuning lifecycle — when fine-tuning beats prompting + RAG, dataset curation, instruction tuning vs preference optimisation (SFT / DPO / RLHF), parameter-efficient methods (LoRA / QLoRA / adapters), evaluation, safety re-tuning, deployment, monitoring, and the cost / benefit framework for choosing between fine-tuning, RAG, and base-model usage.
disable-model-invocation: true
---

# Fine-Tuning Workflows

> **Size budget: 8 KB** — `token-budget.mjs --check`.

Principal-level fine-tuning lifecycle — when fine-tuning beats prompting + RAG, dataset curation, instruction tuning vs preference optimisation (SFT / DPO / RLHF), parameter-efficient methods (LoRA / QLoRA / adapters), evaluation, safety re-tuning, deployment, monitoring, and the cost / benefit framework for choosing between fine-tuning, RAG, and base-model usage.

## Working procedure

1. Establish the relevant mode and existing evidence; do not repeat completed intake.
2. Confirm dataset rights, provenance and privacy; compare held-out evaluation to a baseline before promotion. Record model/data versions, rollback and reload evidence. A training job starting is not a successful promotion.
3. Read only the corresponding sections below before applying their examples.
4. Verify the result with meaningful positive, negative and failure controls. Record actual outcomes, limitations and next action.

## Selected references

- [Purpose](references/procedure.md#purpose) — read when this part of the task applies.
- [Standards Cited](references/procedure.md#standards-cited) — read when this part of the task applies.
- [When to Fire](references/procedure.md#when-to-fire) — read when this part of the task applies.
- [Core Patterns](references/procedure.md#core-patterns) — read when this part of the task applies.
- [Anti-Patterns](references/procedure.md#anti-patterns) — read when this part of the task applies.
- [Verification Checklist](references/procedure.md#verification-checklist) — read when this part of the task applies.
- [Cross-References](references/procedure.md#cross-references) — read when this part of the task applies.
- [Why This Skill Exists](references/procedure.md#why-this-skill-exists) — read when this part of the task applies.
- [Learning hooks](references/procedure.md#learning-hooks) — read when this part of the task applies.

The [full procedure](references/procedure.md) preserves detailed examples and standards.
Load relevant excerpts rather than the entire reference. Runtime capabilities and higher-priority instructions govern imported templates.

Files in this skill

  • SKILL.md2.6 KB
  • references/procedure.md23.9 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…