Run2 json data extraction
[COLM'26] SkillLearnBench is the first benchmark for evaluating continual learning methods that automatically generate agent skills.
npx -y skills add cxcscmu/SkillLearnBench --skill run2_json-data-extractionAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
What its author says it does
Copied from the file, not written here
Deep parsing and structural querying of nested JSON documents, resolving implicit relations across multi-modal corporate datasets like Slack messages and PRs.
SKILL.md
1.5 KB, as published. Nobody here has run it
Advanced JSON Data Extraction Skill
This skill provides refined techniques for extracting, associating, and validating relational data in complex JSON environments, typically found in corporate data exports containing interwoven schemas (e.g., Slack messages, Git PRs, meeting transcripts).
Usage
- Schema Discovery: Use
jqor Python'sjsonmodule to map the root keys (e.g.,['slack', 'documents', 'meeting_transcripts', 'meeting_chats', 'urls', 'prs']). - Text Search Automation: Automate keyword hunting across deeply nested fields (e.g., recursive search for 'strengths and weaknesses' or 'reviewer').
- Cross-Referencing IDs: Trace Employee IDs (
eid_*) discovered in unstructured text (like feedback summaries) or Slack message metadata to resolve entity roles accurately. - URL Validation: Programmatically filter metadata objects (like demo link objects inside a
urlsarray) matching specific competitor names and extract the exact shared properties.
Code Example
import json
def find_competitor_demos(filepath, competitors):
with open(filepath, 'r') as f:
data = json.load(f)
urls = data.get('urls', [])
return [u['link'] for u in urls if any(comp.lower() in u['link'].lower() for comp in competitors)]