transformers

This skill should be used when working with pre-trained transformer models for natural language processing, computer vision, audio, or multimodal tasks. Use for text generation, classification, question answering, translation, summarization, image classification, object detection, speech recognition, and fine-tuning models on custom datasets.

Quality

76%

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 ./backend/cli/skills/llm-tools/transformers/SKILL.md

Quality

Content

72%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.

A well-structured, appropriately short overview that routes to five real reference files, with executable install and quick-start guidance. Its weaknesses are duplicated content between Quick Start and Common Patterns and the near-total absence of validation checkpoints or error-recovery guidance beyond the credential check.

Suggestions

Collapse 'Common Patterns' or differentiate it from Quick Start — Pattern 1 duplicates the pipeline quick start; either show distinct patterns (batch inference, handling gated models) or remove the section and let references/pipelines.md carry it.

Add a brief feedback loop to the credential and model-loading steps, e.g. what to do when a model is gated or $HF_TOKEN is NOT SET, and how to verify a loaded model works before fine-tuning.

Replace placeholder identifiers in the Common Patterns code ('task-name', 'model-id') with one concrete runnable example each, or explicitly justify the placeholders as templates.

Dimension	Reasoning	Score
Conciseness	The body is mostly lean — install commands, tight code blocks, and two-line 'When to use' notes — but the 'Common Patterns' section re-treads ground already covered by Quick Start and the capability sections (e.g., 'Pattern 1: Simple Inference' duplicates the pipeline quick start).	4 / 5
Actionability	Install commands, credential verification, and Quick Start code are copy-paste executable, but the Common Patterns blocks use placeholder identifiers ('task-name', 'model-id') rather than concrete examples, and the fine-tuning pattern omits tokenizer/dataset preparation.	4 / 5
Workflow Clarity	A rough sequence exists (install → verify credentials → quick start → capabilities), and the credential check is a genuine checkpoint, but there are no other validation steps or error-recovery guidance (e.g., gated models, download failures, when a pipeline silently picks a poor default model).	3 / 5
Progressive Disclosure	The body is a clean overview with five one-level-deep, clearly signaled references ('See references/pipelines.md for comprehensive task coverage and optimization'), all of which exist on disk, plus a final index of the reference documentation.	5 / 5
	Total	16 / 20 Passed

Description

80%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.

A strong description with a comprehensive, multi-domain action list and explicit 'should be used when' trigger guidance in third-person voice. Its remaining gaps are missing common synonyms (notably 'Hugging Face' and 'inference') and trigger phrasing that echoes what a user would actually say.

Dimension	Reasoning	Score
Specificity	It lists nine concrete actions spanning NLP, vision, and audio — 'text generation, classification, question answering, translation, summarization, image classification, object detection, speech recognition, and fine-tuning models on custom datasets' — which is comprehensive coverage, matching the top anchor rather than the 'minor gaps' level below.	5 / 5
Completeness	Both halves are present: the what ('working with pre-trained transformer models for natural language processing, computer vision, audio, or multimodal tasks') and the when ('This skill should be used when working with...'). It falls short of the top anchor because the trigger guidance is domain-based rather than user-voice — no 'when the user asks to...' phrasing.	4 / 5
Trigger Term Quality	Good coverage of natural phrases users would say ('text generation', 'question answering', 'summarization', 'fine-tuning'), but common synonyms are missing — 'Hugging Face', 'inference', 'embeddings', 'run a model' — so it stops short of the comprehensive-with-synonyms anchor.	4 / 5
Distinctiveness Conflict Risk	'Pre-trained transformer models' carves a clear niche distinct from document or data skills, but the broad task list (classification, summarization, fine-tuning) overlaps with generic LLM/NLP skills, so it is 'mostly distinct; minor overlap risk' rather than minimal.	4 / 5
	Total	17 / 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
metadata_version	'metadata.version' is missing	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/llm-tools/transformers/SKILL.md
Commit: 4082a2e

Reviewed: 1 day 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.