Automate error analysis and data review on structured datasets.
npx skills add https://github.com/shreyashankar/error-discovery-skill --skill error-discovery
Please help me install this Agent Skill. Skill: error-discovery Source: https://github.com/shreyashankar/error-discovery-skill/tree/main Command: npx skills add https://github.com/shreyashankar/error-discovery-skill --skill error-discovery
This Skill automates error analysis on datasets, providing a visual interface to review and organize failure modes efficiently.
Run the error-discovery skill on your dataset 'traces.jsonl' to automatically analyze errors and visualize the data.
High-intent search queries and answers about installing and using this skill.
To automate error analysis on LLM outputs, run a Python HTTP server-based HTML app that processes structured datasets to identify failure modes. This provides an interactive interface to visualize and organize data quality issues efficiently.
Visually reviewing failure modes is best done through an interactive HTML application served locally via Python. It allows you to cluster, visualize, and review errors within your datasets to spot patterns in AI output quality.
Yes, Python and standard libraries are required to operate the HTTP server and run the visual review app. The server hosts the HTML interface locally, enabling interactive clustering and visualization of your dataset.
Yes, this tool supports data quality assurance on general structured datasets. While it applies to error identification in AI outputs, its automated clustering and visual review capabilities support broader dataset analysis and data review tasks.
Dataset analysis requires clustering and visualization because grouping similar data points reveals underlying failure modes. Visualizing these clusters within an interactive HTML interface makes it easier to review and organize errors that manual inspection might miss.
starlake-aiRun an HTTP server for Starlake deployments with configurable host and port.
maragudkCrafts, reviews, and refines prompts for LLM pipelines, including system and LLM-as-Judge prompts.
breethomasAnalyze LLM pipeline traces to categorize failures and prioritize root causes.
bobmatnycValidate Python data models with Pydantic v2 runtime type checking.
