feat(evals): add coding-task evaluation framework (#2084) by edelauna · Pull Request #2154 · Zoo-Code-Org/Zoo-Code

Commits on Nov 5, 2026

feat(evals): add coding-task evaluation framework (#2084)
edelauna
committed
f69baee
fix(evals): secure coding comparisons and correct grading and cost accounting
edelauna
committed
f49cdb2
fix(evals): enforce trust checks and reliable comparison cleanup and reports
edelauna
committed
95a525c
fix(evals): pass test fixture paths as arguments instead of embedding code
edelauna
committed
bb1758c
fix(evals): enforce isolated grading and safe interruption and result writes
edelauna
committed
12794d7
fix(agent-evals): reject dirty checkouts and clear stale grader digests
edelauna
committed
4f0e44e
fix(agent-evals): prevent child startup after interruption and preserve signal exit codes
edelauna
committed
d26ad55
fix(agent-evals): detect dirty checkouts despite Git settings that hide changes
edelauna
committed
02ce7ff
fix(agent-evals): isolate side builds and invalidate stale results before artifacts
edelauna
committed
980095b
fix(evals): wait for preparation cleanup after repeated interrupts
edelauna
committed
8210012
fix(evals): address review feedback on isolation, cleanup, and reports
edelauna
committed
9028fde
fix(evals): address review feedback on isolation, cleanup, and reports
edelauna
committed
9fd42e9
fix(evals): close review gaps in cleanup and run accounting
edelauna
committed
fceb8a7
fix(evals): bind grader images and isolate host caches
edelauna
committed
6f381c8
fix(evals): clean up partial side publication on failure
edelauna
committed
87901d1
fix(agent-evals): prevent credential leaks to grading subprocesses
edelauna
committed
801c43a
fix(evals): probe container engines and preserve Docker endpoint settings
edelauna
committed
3ce847d

Commits on Nov 6, 2026

Commits on Nov 7, 2026

Commits on Nov 8, 2026

fix(evals): harden grading results and lock standalone side output
edelauna
committed
8f67cd2

Commits on Nov 9, 2026

fix(evals): isolate grader assertions and protect output ownership
edelauna
committed
379aad8

Navigation Menu

Repository navigation

feat(evals): add coding-task evaluation framework (#2084) - #2154

feat(evals): add coding-task evaluation framework (#2084)#2154
edelauna wants to merge 24 commits into
mainfrom
issue/2084

Commits on Nov 5, 2026

feat(evals): add coding-task evaluation framework (#2084)

fix(evals): secure coding comparisons and correct grading and cost accounting

fix(evals): enforce trust checks and reliable comparison cleanup and reports

fix(evals): pass test fixture paths as arguments instead of embedding code

fix(evals): enforce isolated grading and safe interruption and result writes

fix(agent-evals): reject dirty checkouts and clear stale grader digests

fix(agent-evals): prevent child startup after interruption and preserve signal exit codes

fix(agent-evals): detect dirty checkouts despite Git settings that hide changes

fix(agent-evals): isolate side builds and invalidate stale results before artifacts

fix(evals): wait for preparation cleanup after repeated interrupts

fix(evals): address review feedback on isolation, cleanup, and reports

fix(evals): address review feedback on isolation, cleanup, and reports

fix(evals): close review gaps in cleanup and run accounting

fix(evals): bind grader images and isolate host caches

fix(evals): clean up partial side publication on failure

fix(agent-evals): prevent credential leaks to grading subprocesses

fix(evals): probe container engines and preserve Docker endpoint settings

Commits on Nov 6, 2026

fix(evals): qualify grader image and test output permissions

fix(evals): address review feedback and failing CI assertions

Commits on Nov 7, 2026

fix(evals): reuse verified private VS Code cache and preserve proxy settings

fix(evals): disable host commands and lock comparison output

test(evals): verify disabled commands are blocked at dispatch

Commits on Nov 8, 2026

fix(evals): harden grading results and lock standalone side output

Commits on Nov 9, 2026

fix(evals): isolate grader assertions and protect output ownership