Tessl • The management layer for AI agent skills

SkillPublic

roboflow-training-and-evaluation

Use when training Roboflow models, diagnosing why a model underperforms, improving accuracy, or setting up a production feedback loop — covers architecture selection, model IDs, checkpoints, Model Evaluation (confusion matrix, per-class metrics, MCP eval tools), the diagnosis-first improvement playbook, and Active Learning through the Project Model Workflow block.

Quality

88%

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

Quality

Content

81%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-engineered operational skill: routed decision trees, explicit validation and error-recovery loops, and concrete tool-level guidance throughout. The main weaknesses are mild redundancy between the Quick Guide and the decision tree, and inlined reference-grade tables alongside references to bundle files that are not shipped here.

Suggestions

Merge the 'Model Selection Quick Guide' into the decision tree (or reduce it to the table alone) to eliminate the duplicated NAS-guidance prose.

Ship model-ids.md and nas-results.md in the bundle (or move their content into clearly named reference files alongside SKILL.md) so the in-text references resolve.

Consider moving the architecture tables and the MCP tools table into a reference file, keeping SKILL.md to the flow, decision tree, and pointers.

Dimension	Reasoning	Score
Conciseness	The body is dense and assumes competence — architecture tables, a routed decision tree, and error-code handling with no basic-concept padding. Minor over-explanation remains: the 'Model Selection Quick Guide' partially restates the decision tree (it even says 'This summarises the training branches of the decision tree above'), and the NAS section repeats the prerequisites/fallback story twice.	4 / 5
Actionability	Concrete MCP tool names with parameter signatures ('trainings_create(project_id, version_number, model_type="rfdetr-nas-parent")'), exact fallback model IDs, error codes with handling ('nas_not_available_for_plan' → do not retry, fall back to 'rfdetr-medium'), and deep-link URLs. Minor gaps: tool calls lack full example payloads, and the referenced model-ids.md / nas-results.md files are not present in this bundle.	4 / 5
Workflow Clarity	The 14-step decision tree has explicit routing, trial-and-verify checkpoints ('User confirms works → Done. Poor results → Step 13'), prerequisite checks before NAS, and specific error-recovery guidance (treat 'nas_not_available_for_plan' as a plan limit, do not retry, fall back to a named model).	5 / 5
Progressive Disclosure	Good structure with clearly signaled, one-level-deep references ('Read model-ids.md before calling a training tool', MCP resource URIs in Related Pages). However, the large architecture tables and the ~20-line MCP tools reference sit inline in SKILL.md, and the two referenced local files (model-ids.md, nas-results.md) are not present in the bundle.	4 / 5
	Total	17 / 20 Passed

Description

92%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: explicit 'Use when' triggers paired with a concrete inventory of what the skill covers, written in third person without padding. The only gap is keyword coverage — a few natural synonyms a user might say (fine-tune, retrain, evaluate) are absent.

Dimension	Reasoning	Score
Specificity	The description lists multiple concrete capabilities — 'architecture selection, model IDs, checkpoints, Model Evaluation (confusion matrix, per-class metrics, MCP eval tools)', 'Active Learning through the Project Model Workflow block' — with comprehensive coverage of the skill's scope and no vague filler.	5 / 5
Completeness	It explicitly answers both questions: an explicit 'Use when training Roboflow models, diagnosing why a model underperforms, improving accuracy, or setting up a production feedback loop' trigger clause, and a concrete 'covers ...' statement of what the skill contains.	5 / 5
Trigger Term Quality	Natural phrases like 'training Roboflow models', 'diagnosing why a model underperforms', 'improving accuracy', and 'production feedback loop' are present, but common user synonyms such as 'fine-tune', 'retrain', 'model evaluation', or 'mAP' are missing.	4 / 5
Distinctiveness Conflict Risk	'Roboflow models' and 'Model Evaluation' carve out a clear niche with distinct triggers; the only conceivable overlap is with other Roboflow skills (e.g., deployment), which the training/evaluation focus avoids.	5 / 5
	Total	19 / 20 Passed

Validation

93%

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 — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
relative_links	Relative link issues: 4 missing	Warning

	Total	15 / 16 Passed

Repository: roboflow/computer-vision-skills
Path: skills/roboflow-training-and-evaluation/SKILL.md
Revision: ad4f5c6

Reviewed: 7 days ago

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.

roboflow-training-and-evaluation

Table of Contents

Is this your skill?