Waza v0.40.0

Highlights

• Run an exact work unit: Negotiate runner capabilities, validate a
  captured shard/task/trial/attempt and receive an input-derived identity and a
  structured terminal outcome. The mock/text profile runs offline; the
  single-turn Copilot/text profile requires an explicit model and paid-execution
  consent, with tools and skills disabled.
• Use the public runner reference: The Linux/AMD64 image below has been
  anonymously pulled by immutable digest and checked against its signed
  provenance and production-caller mock and bundled native synthetic-provider
  controls. This does not establish live-provider reliability or general
  file/tool/skill support.
• Move a selected AgentV assertion: Pinned offline import/export supports
  one selected case-sensitive contains assertion, not an entire task or
  suite. Unsupported selected semantics fail rather than being approximated.
• Inspect richer evaluation evidence: Task-level MCP mocks, recorded host
  lifecycle events, program-grader context and the Lab conversation viewer
  provide more precise, scoped observations. New results use schema 1.8;
  historical results remain readable.

The public image reference is:

ghcr.io/microsoft/waza-runner-reference@sha256:ded068a636dc721b8e38dc2a68bcbcbdb0d4b33ce5da8d182d562d189a712a39

Its source is commit 93e80d7;
see the publisher
and anonymous-digest verifier.
Only linux/amd64 is published and verified for this image.

Compatibility and operational notes

• A task-level mcp_mocks key that older versions ignored now takes effect.
  before_run also runs before task loading and MCP admission, so review
  existing task files and hooks when upgrading.
• Runner compatibility remains opt-in and limited to one mock/text task, trial
  and attempt. Full external task/suite interoperability remains deferred
  (#789).
• An earlier report of multi-trial timeout/partial-result behavior remains open
  (#865). The new bounded runner
  timeout controls do not establish that the earlier issue is fixed or that it
  reproduces in this release.
• Official standalone binaries report 0.40.0. The immutable tag's version.txt and
  extension.yaml still say 0.39.1 because the established release workflow
  synchronizes those files afterward on main. For a versioned source build use
  make build VERSION=0.40.0; later synchronization does not alter this tag.
• The reference image's binary is built without a release version stamp and
  reports dev; identify this image by its immutable digest and source
  provenance rather than its binary version string.
• The bundled Copilot Go SDK is 1.0.28 and bundled CLI is 1.0.105.

What's Changed

• fix: end responder turns after successful decisions by Ivan (@ivdijan) in #749
• feat(lab): add readable multi-turn conversation transcript by Shayne Boyer (@spboyer) in #753
• chore(deps): consolidate 12 Dependabot updates by Shayne Boyer (@spboyer) in #751
• chore: sync v0.39.1 release registry and documentation by Shayne Boyer (@spboyer) in #748
• chore: consolidate two Dependabot security updates by Shayne Boyer (@spboyer) in #780
• chore(squad): make active roster and routing AI-only by Shayne Boyer (@spboyer) in #781
• chore(squad): add Tess and an every-release media workflow by Shayne Boyer (@spboyer) in #784
• fix: preserve MCP mock sequences across task turns by Erik Galler (@erikgal) in #787
• fix: preserve responder execution errors without a decision by Ivan (@ivdijan) in #796
• feat(runner): add versioned capability negotiation (#757) by Shayne Boyer (@spboyer) in #782
• fix: hide textless messages in Lab conversations by Ivan (@ivdijan) in #801
• chore: make spboyer the sole code owner by Shayne Boyer (@spboyer) in #847
• fix: correct ADC retries, Apps initialization, cancellation, and task diagnostics by Shayne Boyer (@spboyer) in #842
• feat: pass grading context JSON to program graders by Erik Galler (@erikgal) in #799
• ci: harden offline release test gates by Shayne Boyer (@spboyer) in #800
• test: add phase timing to Windows MCP subprocess diagnostics by Shayne Boyer (@spboyer) in #864
• feat(runner): deliver terminal outcomes with deadline precedence #760 by Shayne Boyer (@spboyer) in #793
• feat: record host shell lifecycle and model-turn evidence in results by Erik Galler (@erikgal) in #816
• feat: add task-level mcp_mocks that replace eval mocks by name by Erik Galler (@erikgal) in #797
• docs: align v0.40.0 release readiness guidance by Shayne Boyer (@spboyer) in #869
• feat: add local mock runner container reference by Shayne Boyer (@spboyer) in #871
• feat(runner): execute exact opt-in Copilot text units by Shayne Boyer (@spboyer) in #873
• test(container): verify offline native runner subprocess boundary by Shayne Boyer (@spboyer) in #874
• feat: add pinned AgentV assertion import/export foundation by Shayne Boyer (@spboyer) in #872
• feat(runner): add native-capable runner reference image by Shayne Boyer (@spboyer) in #875
• chore(deps): consolidate current Dependabot updates by Shayne Boyer (@spboyer) in #820

New Contributors

• Ivan (@ivdijan) made their first contribution in #749
• Erik Galler (@erikgal) made their first contribution in #787

Full Changelog: v0.39.1...v0.40.0