What's Changed
- refactor(agents): extract the duplicated agent error logging and attribution by @cyberk1ng in #3837
- feat(demo): give demo tasks descriptions and checklists worth reading by @matthiasn in #3847
- fix(eval): stop the task-agent suite failing correct model behaviour by @matthiasn in #3850
- feat(eval): add a lean-payload variant to probe what limits small models by @matthiasn in #3851
- chore(deps): bump mermaid from 11.16.0 to 11.16.1 in /tool/okf by @dependabot[bot] in #3841
- chore(deps): bump js-yaml from 4.3.0 to 4.3.1 in /docs-site by @dependabot[bot] in #3842
- chore(deps): bump dompurify from 3.4.12 to 3.4.13 in /tool/okf by @dependabot[bot] in #3849
- feat(eval): add the no-op wake, the first scenario that discriminates by @matthiasn in #3852
- feat(agents): the task-agent constitution is code-owned (ADR 0052) by @matthiasn in #3853
- feat(agents): a template's instruction wins, and declares what it overrides by @matthiasn in #3854
- feat(agents): a template's instances become a usable list by @matthiasn in #3855
- test(agents): use the shared GetIt harness, assert both halves of the token claim by @matthiasn in #3856
- fix(settings): the agent one-on-one opens again on desktop by @matthiasn in #3858
- feat(goals): goal-driven agents kickoff — decisions, deterministic core, eval-first harness by @matthiasn in #3857
- feat(goals): goal entities, spec validator, procedural text banners (ADR 0058) by @matthiasn in #3859
- feat(agents): make the 1-on-1 read as one designed surface by @matthiasn in #3860
Full Changelog: 1.0.5+4295...1.0.5+4298