Skip to content

Loki Mode

v7.48.0 Feature

This release adds 3 notable features for engineering teams evaluating rollout.

✓ 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

Summary

AI summary

Updates P3-2, PII-free, and P1-3 across a mixed release.

Full changelog

Fix: shellcheck SC2120 on spec-interrogation.sh (v7.47.0 workflow red)

  • The v7.47.0 spec-interrogation module tripped SC2120 (optional-arg warning) on
    the GH shellcheck runner (a stricter shellcheck version than the local one).
    Added a targeted # shellcheck disable=SC2120 with rationale (the [path]
    argument is optional by design). Tests workflow is green again.

Enterprise: telemetry is now opt-in (P3-2)

  • Telemetry and diagnostics no longer egress by default. Nothing is collected or
    sent unless explicitly enabled (LOKI_TELEMETRY=on or loki telemetry on).
    A default install, including air-gapped / GDPR / FedRAMP environments, sends us
    nothing. All prior opt-out flags still work and always win. Documented the exact
    (PII-free) data inventory in docs/PRIVACY.md.

Verification depth: semantic test-authenticity detector (P1-3)

  • New tests/detect-semantic-test-problems.sh catches fake tests that the
    existing mock/mutation detectors miss: assertions that echo a literal through a
    variable, assertions on a mock's own return value, and net-deleted assertions
    across commits. Conservative by design (never flags a genuine computed
    assertion like expect(add(2,2)).toBe(4)). Available now; wiring it as a
    blocking gate is a tracked follow-up.

Verification depth: real coverage measurement + run manifest (P0-1 Fix A, P3-5)

  • The test gate can now actually MEASURE coverage per language (vitest, jest,
    pytest-cov, go, cargo-llvm-cov) and records it honestly in
    .loki/quality/coverage.json. Measurement is opt-in (it re-runs the suite
    instrumented, so it is off by default to keep autonomous iterations fast): set
    LOKI_COVERAGE_GATE=1 to measure-and-record (warn if below LOKI_MIN_COVERAGE,
    default 80), or LOKI_ENFORCE_COVERAGE=1 to also block a below-threshold-and-
    measurable result. When no coverage tool is present the gate passes through and
    records "not measured" (never fabricates a number).
  • Every run now emits .loki/loki-run.json, a bill-of-materials (spec path+hash,
    model tier used, provider, loki + tool versions, evidence hashes, git SHAs,
    outcome) so a run is auditable and reproducible. Best-effort, never aborts a run.

Each stream built in an isolated worktree, council-reviewed, local-ci green.

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]