Agent skill · Data and AI
eval-engineering
Build evals for an agent repo. Inspect traces, write task specs and audit Harbor tasks.
Build evals for an agent repo. Inspect traces, write task specs and audit Harbor tasks.
- Type
- Agent skill
- Group
- Agent skills
- Category
- Data and AI
- By
- @langchain-ai
- GitHub stars
- 1,316
- License
- MIT
More in Agent skills
Find out what you really want before building: ask, compare approaches, write and approve a design doc.
Read the session transcripts to show why an agent session went wrong, with file and line proof.
Give each independent problem to its own sub-agent and run them side by side.
Carry out a written plan yourself, task by task, with test proof and a ledger, then one final review.
When work is done and green, pick merge, PR or discard, then clean up the branch and worktree.
Handle review comments with checks, not blind yes: understand, verify, then fix or push back.
Read it from code
curl 'https://tools.flisk.si/api/tools/langchain-eval-engineering'