Eval Harness First
by wshobson46891e7e60daNo licenseListed Oct 8, 2026Updated Oct 8, 2026
Build the evaluation harness that gates every fine-tuning run — golden sets, per-failure-mode graders, judge calibration, and base-model baselines. Use when starting a fine-tuning effort, when converting traces into an eval set, or when calibrating a judge against human labels.
Only the file list is public. File contents are available once the skill is installed in a workspace.
| Path | Size | Type |
|---|---|---|
| references/grader-templates.md | 9.9 KB | text/markdown |
| references/judge-calibration.md | 6.5 KB | text/markdown |
| SKILL.md | 7.8 KB | text/markdown |
Source and attribution
Source:wshobson/agentsinplugins/llm-finetuning/skills/eval-harness-firstat commit46891e7
License: No license
Content belongs to its original authors. SourceWeft indexes it from a public repository.
More from wshobson/agents
Uv Package Manager
wshobson
Reference guide for using the uv Python package manager for dependencies, virtual environments and project workflows.
Web Component Design
wshobson
Guides building reusable UI components in React, Vue and Svelte, covering composition patterns, CSS-in-JS choices and component API design.
Python Type Safety
wshobson
Guides Python type annotations, generics, protocols, and strict mypy/pyright checking.
Visual Design Foundations
wshobson
Guides typography, color, spacing and iconography decisions for cohesive, accessible visual design systems.
Responsive Design
wshobson
Guides implementation of responsive web layouts using container queries, fluid typography, CSS Grid and breakpoints.
Python Resource Management
wshobson
Guides Python developers in managing resources deterministically with context managers, cleanup patterns, and streaming.