vision-sft
vision-sft is a Fine-tuning & training data skill for AI agents, published by wshobson in the agents repository on GitHub. Skillspace rates vision-sft C (70/100) and it passed the safety scan; a judge model has also reviewed it.
Fine-tune vision-language models (VLMs) with supervised learning on image+text data. Use when adapting a VLM to a visual domain or task, configuring frozen-vision-tower LoRA, or debugging a VLM fine-tune that trains without learning.
- Grade
- C
- Quality
- 70 / 100
- Safety
- Passed
- GitHub stars
- 37,940
Placed #1 of 6 in the judged category final for Fine-tuning & training data.
This rating describes the version of the repository we crawled; it may have changed since, so go by its current files.
Our review
- Architecture
- 100 / 100
- Maturity
- Structured: ships references or scripts
- Security scan
- Low-risk patterns
- License
- MIT
- Adoption
- Adopted by 0 other repositories, in 0 collections
- Scope
- Portable: works in any project. Generic VLM fine-tuning technique; only references sibling generic skills
Judge review
Placed #1 of 6 in the judged category final for Fine-tuning & training data.
- Trigger
85 85 - Instructions
93 93 - Specificity
93 93 - Safety
88 88 - Disclosure
92 92 - Maintenance
75 75
A mature link in an explicit multi-skill pipeline, with evidence-backed defaults and a genuinely rare debugging section on failure modes that train without erroring or learning.
What the scanner flagged
- mentions secrets in text
Not checked yet
- No blind task eval yet
- Sibling files (scripts, references) not fetched; security scan covers SKILL.md only
Skillspace indexes and rates this skill; it does not host or endorse it. Run third-party skills at your own risk.