Read the results.
Personal context starter checks
Synthetic example. No model was called.
Reports stay in your browser. Checks measure specific rules, not overall model quality.
0/3baseline cases passed3/3context cases passed
Check the answers.
Passing text checks can miss errors.
Test caseWithout contextWith context
project-checkWhich command should I use to check this project?FailPass
Without context
Great question! Try npm test to check your project.
- contains: failed
- notContains: failed
- maxWords: passed
With context
Run `npm run check`.
- contains: passed
- notContains: passed
- maxWords: passed
work-time-zoneWhich time zone should I use for my work calendar?FailPass
Without context
Use the time zone where you work.
- exact: failed
With context
America/Chicago
- exact: passed
status-updateThe first draft is ready. Give me a short status update.FailPass
Without context
The first draft is ready for review.
- contains: failed
- notContains: passed
- maxWords: passed
With context
The first draft is ready. Next: review the result.
- contains: passed
- notContains: passed
- maxWords: passed
Context SHA-256: aa858785865aaa1000253a4796f38d7f3ac02ffb790e24f271fd0065ef3aa42dSuite SHA-256: 9d11e9397f48ba354efc2dc2871e79879f78b3473f7f7c899a3047077298941b
Run it locally.
Choose your files, tasks, and expected results.
- Clone Loooom and open the repository.
- Copy the sample suite. Add your prompts and checks.
- Set the endpoint, model, and key with environment variables.
- Run paired tests. Open the saved report above.
Node.js 22+ · no dependencies
node plugins/context-evals/bin/loooom-context-eval.mjs \
--suite plugins/context-evals/examples/suite.json \
--context ./AGENTS.md \
--output ./context-report.jsonAdd --context ./SOUL.md and --context ./ME.md to compare all three files. Live runs send them to your provider and may incur charges.
Skill repair and ablation tests: Visit the Gym.