Skip to content

Loki Mode

v7.46.0 Breaking

This release includes 4 breaking changes for platform teams planning a safe upgrade.

✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

ai-agents aider anthropic autonomous ci-cd claude
+13 more
cline code-review devops gemini github-action github-issues loki-mode multi-agent openai-codex openapi pull-request-review sdlc spec-driven-development

Affected surfaces

auth rbac

Summary

AI summary

Verification gates are now fully implemented and block as documented, with four new functional gates added and two phantom gates removed.

Full changelog

Verification-credibility sweep (the trust layer is now honest and real)

This release makes Loki's verification layer match its claims. An internal audit
found gates that were advertised but hollow, unwired, or described inaccurately.
The fix: make every gate real or stop claiming it. The deterministic gate set is
now an honest 8.

  • Mock-integrity and test-mutation detectors are now wired as blocking gates.
    tests/detect-mock-problems.sh and tests/detect-test-mutations.sh existed but
    were never invoked by the build loop. They now run every iteration: HIGH-severity
    findings block; Medium/Low are advisory and injected into the next iteration.
    Opt out with LOKI_GATE_MOCK=0 / LOKI_GATE_MUTATION=0. Both detectors honor
    LOKI_SCAN_DIR so they scan the target project, not the Loki install.
  • Anti-sycophancy now acts, not just logs. On a unanimous code-review PASS, a
    Devil's-Advocate reviewer is dispatched; a Critical/High finding blocks
    completion. Previously this path only wrote an audit note. Opt out with
    LOKI_GATE_DEVILS_ADVOCATE=0.
  • Test-coverage gate honesty. The gate ran tests for pass/fail but never
    measured a coverage percentage, while docs claimed ">80% coverage". Docs and
    log labels now state pass/fail accurately; the min_coverage JSON field is
    retained as a target. Real coverage measurement is a planned follow-up.
  • Phantom gates removed. "Input Guardrails" and "Output Guardrails" were
    listed among the gates but had no implementation. Removed. The canonical count
    is now 8: static analysis, test suite (pass/fail), blind 3-reviewer code review
    with severity blocking, anti-sycophancy Devil's Advocate, mock integrity, test
    mutation, documentation coverage, and Magic Modules debate. Backward-compat is a
    conditional healing-mode auditor, not one of the 8.
  • Severity-blocking documented accurately. Code-review blocks on Critical/High
    only; Medium/Low are advisory. Docs and the dashboard-served gate list now match
    the code (previously several surfaces said Medium blocks).
  • Bun-route parity for the new gates and the Devil's-Advocate toggle in
    loki-ts/src/runner/quality_gates.ts.
  • Regression guards added (tests/test-p0-verification-sweep.sh,
    tests/test-p0-gate-behavior.sh, wired into local-ci) so gate-count and
    severity-blocking claims cannot silently drift again.

Reviewed by a 3-of-3 unanimous completion council. No runtime behavior changes
beyond the gates now actually blocking on the conditions documented.

Breaking Changes

  • Removed 'Input Guardrails' and 'Output Guardrails' gates – they were undocumented placeholders.
  • Mock‑integrity detector (`tests/detect-mock-problems.sh`) now runs as a blocking gate by default; opt out with `LOKI_GATE_MOCK=0`.
  • Test‑mutation detector (`tests/detect-test-mutations.sh`) now runs as a blocking gate by default; opt out with `LOKI_GATE_MUTATION=0`.
  • Anti‑sycophancy (Devil's Advocate) reviewer is dispatched and blocks on Critical/High findings; opt out with `LOKI_GATE_DEVILS_ADVOCATE=0`.

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Track Loki Mode

Get notified when new releases ship.

Sign up free

About Loki Mode

Multi-agent autonomous SDLC framework. Spec to deployed app. PRD, GitHub issue, OpenAPI/JSON/YAML, or one-line brief. 5 AI providers, 11 quality gates.

All releases →

Related context

Beta — feedback welcome: [email protected]