Ţéšţš àñđ Éṽàļš
Ŕéšüļţš àŕé ĝŕöüþéđ ƃý ŵĥàţ îš üñđéŕ ţéšţ: éñĝîñé çöŕŕéçţñéšš, ḿöđéļ öüţþüţ àñđ àĝéñţ ƃéĥàṽîöüŕ. Éñĝîñé ţéšţš àššéŕţ éẋþéçţéđ ŕéšüļţš, ḿöđéļ éṽàļüàţîöñš éšţîḿàţé ǫüàļîţý, àñđ àĝéñţ éṽàļüàţîöñš šçöŕé îñđîṽîđüàļ šçéñàŕîöš. Éàçĥ ĝŕöüþ ƒöļļöŵš ţĥé ŕéļéṽàñţ àŕçĥîţéçţüŕé.
7measured
9partial
2unvalidated
16 šţàļé öŕ üñđàţéđ
0/6 ļàýéŕš ŵîţĥöüţ ḿéàšüŕéḿéñţš
Engine and formats3 measured3 partial
Šüƃĵéçţ kapi's own code: what it does to a document it reads, holds and writes back. Éṽîđéñçé Deterministic checks compare the content model and output with expected results. Repeated runs use the same inputs. Ĝàţéš ÇÎ Yes. A failure here stops a build.
6 éṽàļš àçŕöšš F, E · öþéñ →
AI and context3 measured4 partial2 unvalidated
Šüƃĵéçţ What a model writes under kapi's governance, and whether the governance reached it. Éṽîđéñçé Scores estimate performance on the stated corpus. Model-judged scores also depend on the judge's validated agreement with people. Each card reports corpus size to make the sample limitations visible. Ĝàţéš ÇÎ No. Results are tracked over time; regressions require investigation.
9 éṽàļš àçŕöšš C, M, A · öþéñ →
Agent skills1 measured2 partial
Šüƃĵéçţ An agent driving kapi: the shipped Agent Skill, and the MCP tools an assistant calls. Éṽîđéñçé Repeated scenarios measure tool selection and whether the agent completes the task with a passing gate. Scores use observed actions and gate results rather than an LLM judge. Ĝàţéš ÇÎ No. Runs use an interactive agent on a maintainer's machine. CI reads the committed dataset; its date indicates when the results were measured.
3 éṽàļš àçŕöšš S · öþéñ →
Ŵĥàţ ţĥé šţàţüšéš ḿéàñ
measuredruns, data committed, numbers can be read as they standpartialruns, and covers less than the layer needsunvalidatedproduces numbers that should not yet be relied onblockedthe harness runs; the surface it measures returns nothingnot measurednothing measures this