feat: add systematic local Luna eval suites - #325

Merged

jlitola merged 45 commits into

mainfrom

jlitola/daily-agent-evals-metrics

Aug 31, 2026

Commits on Aug 28, 2026

docs: record agent eval phase one merge
jlitola
committed
3802f66
docs: ready local agent eval suites
jlitola
committed
1b2fefd
feat: report logical tool usage
jlitola
committed
d70d728
fix: preserve unknown tool comparisons
jlitola
committed
a9bf87f
feat: add agent eval suite manifest
jlitola
committed
1698e92
feat: separate eval target root
jlitola
committed
af0755f
feat: orchestrate Luna eval suites
jlitola
committed
a0d262a
fix: validate suite artifact identity
jlitola
committed
8713342
feat: add paired suite comparison
jlitola
committed
304ff39
fix: preserve unknown suite tool telemetry
jlitola
committed
a45934a
feat: add named suite CLI
jlitola
committed
d92c7a7
fix: tighten eval suite comparison evidence
jlitola
committed
69286dc
fix: validate imported suite evidence
jlitola
committed
b59efd0
fix: separate dry-run comparison evidence
jlitola
committed
1473b87
test: make eval root paths platform aware
jlitola
committed
59caa3e
docs: record Phase 2 eval validation
jlitola
committed
a03f73f
test: fix suite path portability
jlitola
committed
37c9be1

Commits on Aug 29, 2026

fix: isolate agent eval guidance surfaces
jlitola
committed
5cf74eb
fix: neutralize search source workload prompt
jlitola
committed
10a0142
fix: refresh eval guidance references
jlitola
committed
3e4f386
fix: isolate Codex eval runtime state
jlitola
committed
6c38a2e
refactor: name Codex eval home validator
jlitola
committed
39f850a
fix: preserve GitHits auth for isolated evals
jlitola
committed
4f8f7aa
fix: scope eval home path redaction
jlitola
committed
badbc20
fix: resolve eval guidance paths canonically
jlitola
committed
a00607e
docs: record accepted v4 eval canary
jlitola
committed
283271b
fix: isolate all live Codex eval surfaces
jlitola
committed
4681e12
fix: preserve MCP auth config roots
jlitola
committed
87bc3dd
fix: derive missing MCP config roots
jlitola
committed
f15bb58
fix: handle Windows eval paths and redaction
jlitola
committed
5e8c519
fix: redact escaped paths in eval metadata
jlitola
committed
581fc40

Commits on Aug 30, 2026

feat: add agent eval intent scenarios
jlitola
committed
0967f42
feat: add scenario-aware eval suites
jlitola
committed
51f2ff2
fix: isolate interactive Codex eval sessions
jlitola
committed
8577fce
docs: document agent eval scenarios
jlitola
committed
6562421
fix: enforce historical suite v1 profiles
jlitola
committed
542f4b7
fix: validate interactive Codex eval homes
jlitola
committed
52e42e1

Commits on Aug 31, 2026

docs: record corrected agent eval evidence
jlitola
committed
bcebf1a
fix: expose agent versions in eval reports
jlitola
committed
d895c78
fix: close eval isolation review findings
jlitola
committed
e9a7d4c
docs: record final eval verification
jlitola
committed
5dd00b9
fix: tighten Codex eval home contracts
jlitola
committed
cb5e82a
docs: clarify Codex home isolation wording
jlitola
committed
c7eb650
chore: merge main into daily agent evals
jlitola
committed
fcf9bfb
test: use native eval isolation paths
jlitola
committed
e98562e