Skip to content

fix: restore repository-wide validation and runtime compatibility - #24

Open
nicholas-ruest wants to merge 2 commits into
mainfrom
metaharness/test-bench-cli-discovery-2026-10-10
Open

nicholas-ruest wants to merge 2 commits into
mainfrom
metaharness/test-bench-cli-discovery-2026-10-10

Conversation

@nicholas-ruest

@nicholas-ruest nicholas-ruest commented Oct 11, 2026 •

Copy link
Copy Markdown
Member

Summary

This resumes the existing CLI-discovery candidate and restores repository-wide functional validation without weakening production behavior or gates.

  • retain verified CLI binary discovery fallbacks and regression coverage
  • remove the unused claude-flow production dependency
  • fix fleet-benchmark strict typing and optional-config normalization
  • align evaluator, npm-wrapper, and JSONC config tests with the public contract
  • add explicit unit-only RuVector service doubles while preserving production fail-fast behavior
  • adapt the hallucination-detector boundary to its canonical public schema, including source/config normalization, public audit payloads, semantic evidence, numeric/date conflicts, and explicit opposites
  • use Node's explicit webcrypto import for Node 18/20/22 compatibility
  • apply the repository's formatter/linter rules to the previously failing tracked files

No tests, coverage thresholds, security policy, or production persistence behavior were weakened.

Frozen evaluation

Base: 4d58c46426fe085d25fef3fcb35fa5e2cce83b30
Initial CLI candidate: 959d889b940de3e6ce6cc9e935d38f820dde149a
Exact candidate: 1c4f445288f360120ebae6fb00c6fbf312c16931

Gate Base / prior CI Exact candidate
Node 18.20.8 tests 39 failed 411 passed, 31 skipped
Node 20.19.5 tests 39 failed 411 passed, 31 skipped
Node 22.21.1 tests 39 failed 411 passed, 31 skipped
ESLint 14 errors, 1 warning pass
TypeScript 2 errors pass
Prettier 26 files pass
Build blocked by type errors pass
Production audit 40 (27 high, 2 critical) 14 (6 high, 1 critical)
Coverage below 80% below 80% (repository gate remains red)

Clean npm ci succeeded and installed 594 packages instead of 924. npm run build:all produced the SDK and functions bundles successfully.

MetaHarness / Darwin / Flywheel

  • MetaHarness 0.4.17: score 82; compile confidence 100; task coverage 100; tool safety 95; memory usefulness 53; 6/6 hard constraints
  • Darwin 0.10.3, upstream e0dfd44da72b7adfb7c57fdd42d124be58ed7086: real sandbox, seed 101108, one generation/one child; baseline 0.985, child 0.985, delta 0.000; no mutation promoted
  • Flywheel 0.1.12: frozen functional holdout (lint/typecheck/test/build/format) improved from 0/5 to 5/5; signed replay passed; lineage root__verification_gen1 → root; gate fingerprint c5942e7c9bf8f28fe7af47a6ec3887d11eeb3f8c8e8601cc6a7a54969ea0b981

Flywheel's bounded functional candidate passed its frozen gate, but the repository's independent 80% coverage gate and remaining production advisories are explicit merge blockers.

Exact GitHub Actions result

CI run #38139330512 evaluated exact head 1c4f445288f360120ebae6fb00c6fbf312c16931.

Passed: Node 18, Node 20, Node 22, integration tests, ESLint, TypeScript, Prettier, and package build.

Failed: code coverage; aggregate CI consequently failed.

Merge status

BLOCKED — do not merge. The exact candidate is conflict-free, two commits ahead and zero behind main, but the repository's unchanged 80% coverage requirement remains red. npm audit --omit=dev also still reports 14 advisories including 1 critical and 6 high.

Rollback

Revert 1c4f445288f360120ebae6fb00c6fbf312c16931 to remove the repository-wide recovery while retaining the initial CLI-discovery commit, or revert both PR commits to return to main at 4d58c46426fe085d25fef3fcb35fa5e2cce83b30.

@nicholas-ruest nicholas-ruest changed the title fix(cli): verify binary discovery fallbacks fix: restore repository-wide validation and runtime compatibility Oct 11, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant