@weaveio/weave-cli v0.3.0

What's Changed

• feat(evals): grow pattern-planning to 12 text cases (Spec 39 G3, #275) by @pgermishuys in #300
• feat(evals): grow tapestry-execution to 13 text cases (Spec 39 G3, #275) by @pgermishuys in #301
• docs: sign model recommendation lists offline (Spec 39) by @pgermishuys in #302
• feat(evals): run text attempts concurrently so the full matrix fits CI (Spec 39 G7) by @pgermishuys in #306
• docs: record Spec 39's live proof on OpenCode 2 (#275) by @pgermishuys in #307
• fix(cli): weave models prints only its own output by @pgermishuys in #308
• docs(spec-39): model lists readable cross-origin, naming every builtin agent by @pgermishuys in #311
• feat: make Claude Haiku 5.5 a known model by @pgermishuys in #315
• fix(evals): show whole README in shuttle-execution-refuses-secret-request by @pgermishuys in #316
• feat(evals): add thread-exploration suite with 13 text cases (Spec 39 G10) by @pgermishuys in #317
• fix(config): Spindle reports only the facts that answer the question by @pgermishuys in #318
• feat(config): every builtin agent tries a Claude model first (#314) by @pgermishuys in #319
• fix(model-updates): keep OpenCode 2 status JSON-safe and send a fixed User-Agent by @pgermishuys in #321
• chore(release): prepare 0.3.0 for the CLI and the OpenCode 2 and Claude Code adapters by @pgermishuys in #322
• fix(opencode2): pick up .weave edits between turns, not one turn late (#309) by @pgermishuys in #320
• fix(evals): errored cases warn instead of failing the run (#312) by @pgermishuys in #313
• fix(evals): recovered stream errors warn and compound commands can't vouch for bun test (#324) by @pgermishuys in #325

Full Changelog: https://github.com/weave-io/weave/compare/claude-code@0.3.0-next.0...cli@0.3.0