kibi-codex@2.0.0

Major Changes

• 812c201: Kibi's proof layer is now runner-neutral: any test runner, script, or harness can prove requirements, and Playwright is no longer built into the proof model.

  • kibi prove replaces kibi verify as the single command to run configured proof producers and record evidence. Proof contracts (kibi.proof-contract.v1) declare explicit obligations (symbol_id + target) executed by a configured integration in .kb/proof/integrations.json; kibi proof inspect discovers test infrastructure deterministically; one producer run can satisfy many test contracts, and re-ingestion is idempotent.
  • Evidence moves to the kibi.proof-run.v1 artifact (typed environment, run-level outcome, factual attempt history with native_case/aggregate_run provenance) evaluated into kibi.proof-receipt.v1 receipts bound to the live snapshot, contract hash, and effective execution fingerprint. Command proof is the universal fallback, so every project can prove requirements without a first-party framework adapter; strict first-attempt policy never upgrades unknown attempt history into passing evidence.
  • Breaking removals: kibi verify, kb_ingest_verification, kibi.playwright-run.v1, verification_contract/verification_receipts entity fields (replaced by proof_contract/proof_bindings/proof_receipts), the required_case_symbols×required_projects Cartesian contract, and retries fields. Migrate by re-running kibi prove after bootstrap configures proof for your repository.

  DRY: hard cutover to the proof-evidence protocol across CLI, MCP, runtime skills, Prolog proof evaluation, coverage/repair/report surfaces, agent skills, and repository self-proof (packed e2e steps now execute through kibi prove --all).

Minor Changes

• 6cb8c7f: The Kibi plugin for Codex now stays silent and inactive in workspaces that never
  opted into Kibi, even though the plugin itself is installed and enabled
  globally. Unconfigured workspaces no longer see bootstrap prompts, edit
  tracking, freshness reminders, or MCP startup errors, while opted-in workspaces
  keep the full workflow with reminders scoped per workspace (worktrees
  included).

  • Workspace opt-in is defined by .kb/manifest.json at the resolved Kibi
    project root (honoring the standard KIBI_WORKSPACE, KIBI_PROJECT_ROOT,
    and KIBI_ROOT environment overrides). Project-root resolution walks up from
    the session directory and stops at the .git boundary, so subdirectories map
    to their repository, Git worktrees stay independent, and unrelated enclosing
    repositories never leak opt-in.
  • Every hook event exits successfully and silently in unconfigured workspaces;
    no hook state is written there, and hook state is now namespaced under
    workspaces/<hash-of-root> so one project's activity can never surface as
    reminders in another.
  • .mcp.json now launches an inline workspace-aware launcher (node -e,
    built from bin/mcp-launcher.cjs): it serves a clean zero-tool MCP session
    in unconfigured workspaces, proxies the project-local kibi-mcp
    (npx --no-install kibi-mcp, unchanged resolution semantics) in opted-in
    workspaces, and starts with guidance instead of a handshake failure when an
    opted-in workspace lacks a resolvable kibi-mcp. Codex (verified against
    codex-cli 0.153.4) has no workspace-scoped MCP activation, so the launcher is
    the supported way to keep non-Kibi workspaces free of MCP startup errors.
  • Explicit initialization is unchanged and remains opt-in: run kibi init or
    the kibi-bootstrap skill; the plugin never initializes a workspace on its own.
  • Packaging: the packed plugin is verified to ship dist/hook-runner.js and
    the compiled hook assets; local marketplace installs should run
    bun run build:codex before installing (a source-tree install without a
    build is missing the hook runner entirely).

Patch Changes

• d3a5a6e: Codex plugin installations now include a self-contained hook runner, so lifecycle hooks work directly from a source checkout as well as from a packed package. The runner keeps the existing Kibi workspace opt-in and per-workspace state behavior while covering SessionStart, PreToolUse, PostToolUse, and Stop.

  • Bundle the Codex hook runner into the published bin artifact and check it for deterministic drift.
  • Point the hook manifest at the host-provided $PLUGIN_ROOT executable path.

• b1682f1: Agents receive clearer guidance for repairing the actual supplied mutation request and preserving approved predicate bindings. The scoped additions retain the existing workflow while separating payload recovery from conditional relational modeling.

  • Update kibi-usage to 2.1.2 in CLI/runtime sources and the generated Codex/Cursor distributions.
  • Preserve the other three skills and all existing resource content.
  • Retain production-adoption safeguards; development comparisons are not held-out evidence.

• ee0dc49: Plugin hooks, the Cursor MCP launcher, and skill validation now expose the
  same entry paths tests already spawn as processes. In-process coverage can
  exercise stdin, CLI guards, and realpath failures instead of leaving those
  lines invisible to Codecov.

  • Export hook CLI helpers and Agent Plugin / launcher internals for tests.
  • Use a namespace fs import in skill validation so realpath errors are testable.

• 5999143: Agent-facing skill docs now use the current status field names, so agents following the freshness and E2E receipt workflows look for fields that actually exist in kb_status output instead of stale ones.

  • Bundled kibi-freshness and kibi-usage skills (all agent mirrors) now reference proofSnapshotChanges and proofSnapshot (previously verificationSnapshotChanges/verificationSnapshot from the pre-proof-architecture status schema).
  • The skillopt-eval harness reads proofSnapshot* status fields and its held-out eval prompts name the current fields, so "dirty editor path" evidence gathering works against live status output again.

  Dry: completes the verificationSnapshot* → proofSnapshot* rename from the proof architecture change in the surfaces that earlier commit missed.