perf(tooling): adopt cooperative fast local validation

[gfyrag]

Summary

• replace per-run Go/module/lint/XDG cache trees with stable shared external caches under ~/.cache/ledger-ai, retaining only run-local temporary state
• split validation into a fast compile/invariant baseline, an exact-diff targeted PR selector, and an explicit manual full fallback
• remove duplicate normalization/pre-commit replays from review and adoption while preserving exact candidate review, target freshness, and leased push gates
• document the cooperative-agent threat model and CI as the authoritative broad clean-validation boundary

Decision record

INITIAL_TARGET_BASE_SHA: 0d5cd3f619915e8c53a47559eac5fa31d1f75e82

PUBLICATION_BASE_SHA after #1870: 6d76d60adde8159b66a98c12f3e49757e636a985

Reference measurements with the pinned toolchain:

• isolated cold pre-commit: over 10m31s, ending in the operator lint timeout
• shared warm full pre-commit: 64.83s
• fast baseline: 22.44s
• exact tooling PR profile: 2m26s before fix(test): isolate nested AI tooling fixtures #1870; 3m56s with fix(test): isolate nested AI tooling fixtures #1870's expanded fixture tests
• warm full fallback: 7m32s end to end
• concurrent lint in two detached worktrees sharing caches: 35.97s each, zero findings

The simple warm pre-commit was selected over #1869's roughly 1,500-line dependency selector. #1862's verified seed/copy module-cache design is classified SUPERSEDED_BY_SIMPLER_CACHE_MODEL. Merged #1866 same-run exact-state reuse is preserved. The local proportional-validation branch is superseded by the smaller conservative selector. #1870 merged during this PR and now supplies the canonical accidental parent-binding sanitizer; the rebase retained it and removed the redundant selector-level scrub. #1864 has no design collision.

Local / CI split

Local always: repository invariants, root compilation, diff hygiene, exact review, base/head freshness, and guarded publication.

Local when affected: pre-commit normalization, focused race/regression tests, E2E, scenarios, Schemathesis, and operator tests.

CI by default: full root race/coverage, broad E2E/model/scenario/Schemathesis/operator execution, and Antithesis workload.

Validation

• AI_REVIEW_BASE_SHA=6d76d60adde8159b66a98c12f3e49757e636a985 bash scripts/agent-check-pr
• shared-cache cooperative scenarios: same source, changed source, build tags, race/non-race, and concurrent builds
• current pinned golangci-lint in two concurrent detached worktrees
• shell syntax and diff whitespace checks

Broad clean validation is intentionally delegated to this PR's independent CI.

[NumaryBot]

✅ Approve — automated review

No actionable correctness defects were identified in the current diff. The prior review had no active findings to revalidate.

No findings.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 77.03%. Comparing base (6d76d60) to head (77f32f8).

Additional details and impacted files

@@               Coverage Diff                @@
##           release/v3.0    #1871      +/-   ##
================================================
+ Coverage         77.01%   77.03%   +0.01%     
================================================
  Files               477      477              
  Lines             51141    51141              
================================================
+ Hits              39386    39395       +9     
+ Misses             8320     8313       -7     
+ Partials           3435     3433       -2     

Flag	Coverage Δ
e2e	77.03% <ø> (+0.01%)	⬆️
scenario	77.03% <ø> (+0.01%)	⬆️
unit	77.03% <ø> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.