llm-eval
Skill1 starsChanged 5 months agoMITView on GitHub
Build or run a ground-truth accuracy harness for an AI/LLM feature, measuring against an authoritative reference with a pass threshold. Use when adding, changing, or validating an LLM-backed extraction/classification/generation feature.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- LuisSambrano/antigravity-configtheir own version, first seen Nov 8, 2026
- digitaldreams/tuhintheir own version, first seen Nov 1, 2026
- BrendenKennedy/claude-for-ai-platformstheir own version, first seen Oct 13, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- LuisSambrano/antigravity-config★ 8same name
- digitaldreams/tuhin★ 3same name
- BrendenKennedy/claude-for-ai-platforms★ 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 repository4 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- cc-task-benchSkill
- tune-parametersSkill
- verify-refactorSkill
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 0 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 15
Keep the harness provider-agnostic (call the project's LLMProvider interface, not a vendor SDK directly) so the same eval runs across …
“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
Goal: never ship an AI feature on vibes — measure it against known-correct data.
- Find an authoritative reference. A public dataset, a hand-curated gold set, or a deterministic oracle. Without ground truth there is no eval — say so and ask the user for a source.
- Build the dataset. A list of
{ input, expected, tolerance? }cases drawn from that source. Include edge cases (empty, malformed, ambiguous), not just the easy ones. - Run the feature over every case. For numeric outputs, compare within per-field
tolerance; for structured/text outputs, compare with the right notion of equality (exact, set, semantic). - Gate on accuracy. Compute
passed / total, log it, and assert a threshold (e.g.>= 0.90). Wire this into the test suite so regressions fail CI. - Log inputs and outputs during runs (
[LLM] Input:/[LLM] Response:) so failures are debuggable and prompt changes are measurable over time.
Keep the harness provider-agnostic (call the project's LLMProvider interface, not a vendor SDK directly) so the same eval runs across models without re-instrumentation.
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.