Commits

History for

onv0.1.0

Commits on Aug 16, 2026

harden: close every path that could return PASS without checking

ChelseaKR
and
claude
committed
5caf8e5
couplings: intersect the shared cause over the suites that failed, not every enabled one

ChelseaKR
and
claude
committed
15a434a
report: say when three failures are one finding

ChelseaKR
and
claude
committed
5c8c184
suites: passage_attribution, the check that can say "wrong paragraph"

ChelseaKR
and
claude
committed
e798d13
bundle: an item may declare which passage actually answers it

ChelseaKR
and
claude
committed
dbf81ba
self: hold Plumbline to the standard it holds its targets to

ChelseaKR
and
claude
committed
0695bef
adapters: record against a program, with no socket anywhere in the run

ChelseaKR
and
claude
committed
e483166
proof: a defect-injection matrix, because thirteen green ticks prove nothing

ChelseaKR
and
claude
committed
a565b21
demo: grow the bundle to 174 items so the statistics discriminate

ChelseaKR
and
claude
committed
f7c9fd7
languages: recognise Arabic by script, and let a target declare its own

ChelseaKR
and
claude
committed
1b92472
judges: warn when a live model judge has nowhere to record what it decided
ChelseaKR
committed
495fdf9
validate: accept a question set, and say what the evidence actually is
ChelseaKR
committed
bba5a0c
judges: an optional model judge that cannot quietly weaken a report
ChelseaKR
committed
92182d6
adapters: record from a live target, without letting the gate near it
ChelseaKR
committed
7071783
net: one bounded module for every outbound call, and a test that keeps it that way
ChelseaKR
committed
8e9a6d3

Commits on Aug 15, 2026

gate: a CI entry point and the pin file a consuming repo copies

ChelseaKR
and
claude
committed
e84595a
baseline: regression comparison that refuses to compare incomparable runs

ChelseaKR
and
claude
committed
4ea6309
demo: extend the riverbend bundle to exercise all thirteen suites

ChelseaKR
and
claude
committed
ab14a0e
suites: accessibility, computed from a captured interface snapshot

ChelseaKR
and
claude
committed
559c634
suites: multilingual fidelity, adversarial, fairness, harms, privacy

ChelseaKR
and
claude
committed
a75cc33
suites: groundedness, citation validity, citation accuracy

ChelseaKR
and
claude
committed
77cca32
suites: cross-language agreement, the check pooled averages cannot make

ChelseaKR
and
claude
committed
f1b15fe
stats: confidence intervals and minimum detectable effect on every suite

ChelseaKR
and
claude
committed
d5e27ef
chore: gitignore bytecode caches; drop __pycache__ committed by mistake

ChelseaKR
and
claude
committed
8b27745
tests: hashing, bundle integrity, suites, and end-to-end CLI exit codes

ChelseaKR
and
claude
committed
86dd70d

Navigation Menu

Commits

Commits on Aug 16, 2026

harden: close every path that could return PASS without checking

couplings: intersect the shared cause over the suites that failed, not every enabled one

report: say when three failures are one finding

suites: passage_attribution, the check that can say "wrong paragraph"

bundle: an item may declare which passage actually answers it

self: hold Plumbline to the standard it holds its targets to

adapters: record against a program, with no socket anywhere in the run

proof: a defect-injection matrix, because thirteen green ticks prove nothing

demo: grow the bundle to 174 items so the statistics discriminate

languages: recognise Arabic by script, and let a target declare its own

judges: warn when a live model judge has nowhere to record what it decided

validate: accept a question set, and say what the evidence actually is

judges: an optional model judge that cannot quietly weaken a report

adapters: record from a live target, without letting the gate near it

net: one bounded module for every outbound call, and a test that keeps it that way

Commits on Aug 15, 2026

gate: a CI entry point and the pin file a consuming repo copies

baseline: regression comparison that refuses to compare incomparable runs

demo: extend the riverbend bundle to exercise all thirteen suites

suites: accessibility, computed from a captured interface snapshot

suites: multilingual fidelity, adversarial, fairness, harms, privacy

suites: groundedness, citation validity, citation accuracy

suites: cross-language agreement, the check pooled averages cannot make

stats: confidence intervals and minimum detectable effect on every suite

chore: gitignore bytecode caches; drop pycache committed by mistake

tests: hashing, bundle integrity, suites, and end-to-end CLI exit codes

Breadcrumbs

User selector

Datepicker

Commit history

Commits on Aug 16, 2026

Commits on Aug 15, 2026