Name: tao-run-deft-aoi
Rating: 70.39999999999999 (1 reviews)
Author: NVIDIA

tao-run-deft-aoi

Run the full DEFT AOI improvement loop for NVIDIA TAO VisualChangeNet / ChangeNet PCB inspection models: baseline evaluate, RCA, Cosmos AnomalyGen / AMP synthetic defects, k-NN mining, retraining, and deployment gating until FAR / recall KPI targets are met. Use for prompts like "run the DEFT loop", "fine-tune until FAR below 0.1% at recall=100%", or "improve my AOI ChangeNet model with RCA and synthetic defects"; do not use for standalone TAO training, one-off inference, generic anomaly generation, or RCA-only analysis.

Quality

86%

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.

The body is a well-structured orchestration overview with strong workflow clarity (explicit sequence, validation, and recovery loops) and good progressive disclosure through a reference table. Its weak spots are minor cross-file redundancy (AutoML pitfall, path rule) and a broken bundle reference to the missing agents/reporter.md.

Suggestions

Add the missing agents/reporter.md file (or correct the path) — it is referenced as the only delegated subagent in the loop but the agents/ directory does not exist in the bundle.

De-duplicate the AutoML-policy pitfall and the path-rule invariant: state them once in SKILL.md and have the reference files link back, rather than restating each in both places.

Consider inlining one concrete docker run / script invocation example in the body for the most common stage so the body alone is copy-paste ready for the happy path, keeping detailed variants in references.

Dimension	Reasoning	Score
Conciseness	The body is dense and operational with no concept-padding Claude already knows, but the AutoML-policy pitfall and the path rule are each restated in both SKILL.md and a reference file — minor redundancy that could be trimmed, placing it above the score-3 anchor but short of fully lean.	4 / 5
Actionability	Concrete scripts with argument signatures, pinned env-var exports, exact path rules, hard-stop conditions, and state-file schemas give mostly executable guidance; full copy-paste docker commands are deliberately deferred to references (good disclosure), so the body alone has minor gaps versus the score-5 'copy-paste ready' anchor.	4 / 5
Workflow Clarity	A clearly sequenced Pre-Flight → Baseline → Iterate → Stop → Render pipeline with explicit validation checkpoints (mid-iteration + post-assembly leakage checks, CSV existence check, post-stage disk re-read), feedback loops (blocker recovery: fix → resume step N), and hard-stop gates — matching the score-5 anchor; the destructive/batch validation cap does not apply because validation is extensive.	5 / 5
Progressive Disclosure	Clear overview with a navigation table and one-level-deep references appropriately split across references/*.md and scripts/, but agents/reporter.md — referenced repeatedly as the only delegated subagent in the loop — is missing from the bundle, a navigation gap that keeps it below the score-5 anchor.	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.

The description is specific, complete, and distinctive: it enumerates the concrete loop stages, gives natural trigger phrases for when to use it, and negative guidance for when not to. Its only mild gap is trigger-term breadth (no file extensions / synonyms), which keeps trigger_term_quality at 4 rather than 5.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — 'baseline evaluate, RCA, Cosmos AnomalyGen / AMP synthetic defects, k-NN mining, retraining, and deployment gating' — covering the full loop comprehensively, matching the score-5 anchor.	5 / 5
Completeness	Explicitly answers both what ('Run the full DEFT AOI improvement loop…') and when ('Use for prompts like…') with concrete trigger phrases, plus negative guidance — the score-5 anchor match.	5 / 5
Trigger Term Quality	Natural trigger phrases are quoted directly ('run the DEFT loop', 'fine-tune until FAR below 0.1% at recall=100%', 'improve my AOI ChangeNet model with RCA and synthetic defects'); coverage is good but not exhaustive — no file extensions and a few synonyms missing, so it sits above the score-3 anchor but below 5.	4 / 5
Distinctiveness Conflict Risk	Clear niche (DEFT AOI loop on TAO VisualChangeNet PCB inspection) with explicit 'do not use for standalone TAO training, one-off inference, generic anomaly generation, or RCA-only analysis' minimizing conflict with sibling skills.	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
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: NVIDIA/skills
Path: skills/tao-run-deft-aoi/SKILL.md
Commit: be3d1a4

Reviewed: 4 days 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.