Skip to content

Loki Mode

v7.51.0 Breaking

This release includes 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

Summary

AI summary

Coverage artifact honesty now writes coverage.json with measured:false when off and LSP‑diagnostics gate is fully implemented on both routes.

Full changelog

Tech-debt closure: four inert/half-wired features made real (both routes)

This release removes scaffolding that read like a feature but did nothing,
honoring the no-tech-debt mandate. Each activation is advisory-first and
deny-filters clean results, so an activated gate can never false-fire and
deadlock the autonomous loop. Defaults are unchanged for existing users.

  • Coverage artifact honesty (autonomy/run.sh): at the default (measurement
    off) the coverage block was skipped entirely, so no coverage.json was ever
    written -- a missing-artifact gap in the reproducibility manifest. The gate now
    writes coverage.json with measured:false even when measurement is off, with
    zero added runtime (no instrumented re-run) and no blocking. Measurement stays
    opt-in (LOKI_COVERAGE_GATE=1); default-on would double every test run.
  • LSP-diagnostics gate made real on both routes: the gate read
    .loki/quality/lsp-diagnostics.json but nothing wrote it (inert). A
    route-neutral writer (mcp/lsp_proxy.py --write-diagnostics) is now invoked
    identically from the Bun gate (loki-ts/src/runner/quality_gates.ts) and the
    bash gate (autonomy/run.sh), with byte-identical blocking semantics: single
    knob LOKI_GATE_LSP_DIAGNOSTICS (default off); count_errors > 0 blocks only
    when enabled; absent/malformed/clean never fires; no language server present ->
    writes nothing (never fabricates a clean verdict). LOKI_GATE_LSP_WRITER=0
    lets a caller supply a pre-built artifact. The bash/Bun parity test now
    validates these as genuinely shared toggles (prior bun-only carve-out removed).
  • Evidence-gate-details consumer (autonomy/run.sh, closes P1-1): the
    completion council wrote .loki/council/evidence-gate-details.json but run.sh
    had no consumer. run.sh now surfaces the details advisory (WARN on block, INFO
    otherwise) without introducing a new block; the wrapper preserves the gate's
    exact return code, so completion detection is unchanged. Absent/malformed file
    degrades silently.
  • Approval phase-gate enforcement (autonomy/run.sh): a policy
    "approval required" result (exit 2) previously logged and proceeded. It now
    honors an approval wait on .loki/signals/POLICY_APPROVED|POLICY_REJECTED
    (same mechanism as staged autonomy) when LOKI_STAGED_AUTONOMY=true or
    LOKI_POLICY_APPROVAL_ENFORCE=1. With neither knob set, behavior is unchanged
    (advisory log + proceed).

Gates: local-ci 78/78, full pytest + bun test + bash/Bun parity matrix green,
3-reviewer council unanimous APPROVE (2 Opus adversarial + 1 Sonnet).

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]