model-merging

Merge multiple fine-tuned models using mergekit to combine capabilities without retraining. Use when creating specialized models by blending domain-specific expertise (math + coding + chat), improving performance beyond single models, or experimenting rapidly with model variants. Covers SLERP, TIES-Merging, DARE, Task Arithmetic, linear merging, and production deployment strategies.

Quality

82%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

The canonical home for this skill is model-merging in OpenLAIR/dr-claw

Quality

Content

65%

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

A thorough, highly actionable skill with excellent executable configs, undermined by significant cross-section duplication and inline method detail that overlaps with the provided references. A single sequenced merge→validate→deploy workflow with an explicit benchmarking checkpoint would raise workflow clarity.

Suggestions

Collapse the duplicated Linear/SLERP/TIES/DARE configs from 'Core Concepts', 'Merge Methods Guide', and 'Best Practices' into one canonical section, offloading deep method detail to references/methods.md.

Replace the implicit pitfall with an explicit numbered workflow (write config → run merge → benchmark on MMLU/HumanEval/GSM8K → only deploy if metrics hold) that includes a validation checkpoint.

Move the dated 'best on Open LLM Leaderboard 02/2024' success claim into a deprecated/old-results section or drop it, and reference evaluation.md inline at the evaluation step rather than only in 'See Also'.

Dimension	Reasoning	Score
Conciseness	Mostly actionable but padded by duplication — Linear/SLERP/TIES/DARE configs reappear across 'Core Concepts', 'Merge Methods Guide', and 'Best Practices', plus math formulas (SLERP, task vectors) and a dated '02/2024' leaderboard claim that Claude largely already knows.	2 / 3
Actionability	Provides complete, copy-paste-ready mergekit YAML configs, bash run commands, and python eval/upload snippets, matching the score-3 anchor of fully executable, specific examples.	3 / 3
Workflow Clarity	Steps exist (install → config → merge → evaluate → deploy) and evaluation is encouraged, but there is no explicit validation checkpoint in the merge-to-deploy sequence; the risky 'deploy without testing' path is only addressed as a pitfall, capping clarity at 2.	2 / 3
Progressive Disclosure	Three real one-level-deep references (methods.md, examples.md, evaluation.md) are listed under 'See Also', but they are signaled only at the very end while substantial method-detail that could live in references is kept inline, matching the score-2 anchor of structure with content that should be separate kept inline.	2 / 3
	Total	9 / 12 Passed

Description

100%

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

A strong, third-person description that states concrete capabilities, names the supported methods, and gives an explicit 'Use when' trigger clause covering several natural scenarios. It avoids fluff and is clearly distinct from other skills.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions in third person ('Merge multiple fine-tuned models', 'combine capabilities without retraining', 'blending domain-specific expertise', 'production deployment strategies'), matching the score-3 anchor of multiple specific concrete actions.	3 / 3
Completeness	Explicitly answers both 'what' and 'when' via the 'Use when creating specialized models..., improving performance..., or experimenting rapidly...' clause, matching the score-3 anchor with explicit triggers.	3 / 3
Trigger Term Quality	Covers natural terms a user would say ('fine-tuned models', 'mergekit', 'specialized models', 'model variants') plus the method names, giving good coverage of natural trigger language rather than only jargon.	3 / 3
Distinctiveness Conflict Risk	The model-merging niche with mergekit-specific triggers (SLERP, TIES-Merging, DARE, Task Arithmetic) is clearly distinct and unlikely to fire for unrelated skills.	3 / 3
	Total	12 / 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
skill_md_line_count	SKILL.md is long (541 lines); consider splitting into references/ and linking	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/ml-training/model-merging/SKILL.md
Commit: 52845c3

Reviewed: about 9 hours 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.