feat(metrics): ST1 — loop metrics + adversarially-hardened FCR/RP baseline#16

Merged

SollanSystems merged 15 commits into

mainfrom

feat/st1-metrics-baseline

Jul 4, 2026

Commits on Jul 3, 2026

feat(schemas): canonical repair-record + rollout-record schemas (ST1 AC1)
SollanSystems
committed
feat(metrics): derive FCR/RP from real .loop evidence with recompute-and-reject (ST1 AC2-AC5)
SollanSystems
committed
feat(rollout-ledger): reject-on-disagree via shared recheck_productive (ST1 AC2)
SollanSystems
committed
feat(contract): validate repair/rollout/receipt records when present (ST1 AC6)
SollanSystems
committed
feat(cli): wire 'loop metrics' subcommand + console-script entry (ST1 Rider A/B)
SollanSystems
committed
docs(eval-suite,example): name canonical repair record; relocate example record to .loop/repair (ST1 AC1)
SollanSystems
committed
feat(metrics): publish real FCR/RP baseline + README Metrics passage (ST1 AC5)
SollanSystems
committed
fix(metrics): close ST1 evidence/FCR/RP gaming gaps

SollanSystems
and
claude
committed
fix(contract): report validated record schemas in schemas_checked

SollanSystems
and
claude
committed
refactor(rollout): rename summarize key to rollout_productivity

SollanSystems
and
claude
committed
docs: canonical repair-record path, evidence-not-proof, README baseline binding

SollanSystems
and
claude
committed
chore(metrics): regenerate gate-backed baseline scorecard

SollanSystems
and
claude
committed
fix(metrics): outcome-class FCR, verdict-only baseline, task-keyed RP anchoring (round-2 adversarial findings)

SollanSystems
and
claude
committed
chore(metrics): restamp baseline at the round-2-hardened derivation

SollanSystems
and
claude
committed

Commits on Jul 4, 2026

fix(tests): make live-contract and foreign-cwd tests fresh-checkout-safe

SollanSystems
and
claude
committed