- Home
- Skills
- Workflow Automation
- Dataset Label QC System
More screenshots
Works with the AI Agents you already use
Dataset Label QC System
Catch label errors before training, with agreement checks by class and targeted re-review ordered by impact.
$5
Dataset Label QC System
Example session with this skill installed
Three double labels: one disagreement, two agreements.
- Read your context and instructions
- Compiled the dataset label qc
Verbatim script output on the shipped sample:
n=3 overall_agreement=66.7%
class intent-a: agreement=50.0% n=2
class intent-b: agreement=100.0% n=1
conflict intent-a vs intent-b: 1
Reading: overall agreement hides the weak class; intent-a agrees only 50.0 percent of the time. The full worked example in SKILL.md walks through the same method on a larger realistic case, including the limits of the first fix.
Connects securely to your tools. The creator never sees your data.
About this skill
Dataset Label QC System is a step-by-step workflow for labelled dataset quality control for training and evaluation data. It ranks causes by measured impact before any fix is attempted, so effort goes to the largest contributor first. It includes diagnostic trees, thresholds to state in advance, decision tables, templates, checklists, and a scoring rubric, plus a runnable script that performs the core ranking or measurement on real input. The worked example in the skill shows the method end to end on a realistic case, including what remains unresolved after the first fix.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Workflow Automation.
Results depend on the quality and coverage of the input data; missing logs, missing timings, or a biased sample will limit conclusions. Thresholds in the skill are defaults and must be adjusted to the system at hand. The script performs the core measurement only; judgement about causes, guards, and rollout remains with the user.
Agreement measures consistency, not correctness; two labellers can agree and still be wrong, which is why the skill requires adjudication of the worst pairs and a fresh-sample re-measure.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 23 days ago