eval-driven-development
Skill21 starsChanged 5 months agoMITView on GitHub
Building rigorous LLM evaluation pipelines. Use when developing AI features to ensure quality and prevent regressions across different model versions.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- inferedge-ai/evalkittheir own version, first seen Nov 1, 2026
- anandamritraj/goaltheir own version, first seen Sep 15, 2026
- ContextJet-ai/awesome-llm-observabilitytheir own version, first seen Jul 6, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- ContextJet-ai/awesome-llm-observability★ 46same name
- inferedge-ai/evalkit★ 14same name
- anandamritraj/goal★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository35 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- api-and-interface-designSkill
- ax-extract-workflowSkill
- browser-testing-with-devtoolsSkill
- ci-cd-and-automationSkill
- code-review-and-qualitySkill
- code-simplificationSkill
- context-engineeringSkill
- curlshotSkill
- debugging-and-error-recoverySkill
- deprecation-and-migrationSkill
- documentation-and-adrsSkill
- doubt-driven-developmentSkill
- frontend-ui-engineeringSkill
- git-workflow-and-versioningSkill
- hermes-tweetSkill
- idea-refineSkill
- incremental-implementationSkill
- interview-meSkill
- linkedinSkill
- loootSkill
- mcp-tool-integrationSkill
- observability-and-instrumentationSkill
- performance-optimizationSkill
- plan-ledger-tasks-yyloSkill
- planning-and-task-breakdownSkill
- security-and-hardeningSkill
- shipping-and-launchSkill
- source-driven-developmentSkill
- spec-driven-developmentSkill
- test-driven-developmentSkill
- using-awesome-ai-skillsSkill
- video-routerSkill
- xquik-public-dataSkill
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 17
Curate Golden Dataset: Create diverse test cases including edge cases.
- Distinctive: line 33
No golden dataset exists
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Eval-Driven Development
Overview
Eval-Driven Development ensures that AI features behave deterministically and predictably by testing them against a golden dataset using automated evaluators.
When to Use
- Building an AI-powered feature
- Tuning prompts or changing underlying models
- Implementing RAG pipelines
Process
- Curate Golden Dataset: Create diverse test cases including edge cases.
- Define Metrics: Choose appropriate evaluators (e.g., exact match, semantic similarity, LLM-as-a-judge).
- Run Pipeline: Execute the AI feature over the dataset and collect results.
- Analyze Failures: Inspect low-scoring examples and update prompts or logic.
- Establish Baseline: Set a minimum threshold for CI/CD checks.
Common Rationalizations
| Rationalization | Why It Is Wrong |
| "Manual spot checks are enough." | Spot checks miss regressions across prompts, model versions, and edge cases. |
| "We can add evals after launch." | Without a baseline, you cannot tell whether a later prompt or model change improved behavior. |
| "The judge model says it is good." | LLM judges need criteria, calibration examples, and failure review before they are trustworthy. |
Red Flags
- No golden dataset exists
- Metrics are vague or not tied to user-visible quality
- Low-scoring examples are ignored instead of inspected
- The baseline threshold is chosen after seeing the desired result
Verification
Before finishing, confirm:
- The golden dataset includes normal, edge, and known-failure cases
- Metrics and evaluator prompts are committed or otherwise reproducible
- The current model/prompt has a recorded baseline
- Failure examples have been reviewed and categorized
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.