.omo/evidence/20260704-fix-5788/qa-summary.md
PATH="$HOME/.bun/bin:$PATH" bun test packages/model-core after adding only the claude-sonnet-5 assertion.PATH="$HOME/.bun/bin:$PATH" bun .omo/evidence/20260704-fix-5788/resolve-sonnet-limit.mts.PATH="$HOME/.bun/bin:$PATH" bun test packages/model-core.PATH="$HOME/.bun/bin:$PATH" bun run typecheck.bun run packages/shared-skills/skills/programming/scripts/typescript/check-no-excuse-rules.ts ... and pure LOC measurement for the touched TypeScript files.red-bun-test-packages-model-core.txt: the new regression failed before product code changes with Expected: 1000000 and Received: 200000.before-resolver-output.txt: anthropic/claude-sonnet-5 actualLimit=200000.after-resolver-output.txt: anthropic/claude-sonnet-5 actualLimit=1000000.green-bun-test-packages-model-core.txt: 284 pass, 0 fail, exit_code=0.typecheck.txt: full repo typecheck exited 0.no-excuse-rules.txt: no violations in the three touched TypeScript files.loc-check.txt: touched files are 40, 131, and 7 pure LOC.200000 to 1000000 for the affected provider/model pair.opencode-qa was consulted: no OpenCode process, hook, server, TUI, installer, or DB surface is changed. A broader OpenCode harness run would not exercise additional behavior beyond this pure model-core resolver path.script/agent/setup.sh was run first as requested, but its build phase failed because strict Codex plugin materialization retried local file submodule transport for shared-skill upstreams. The failure was scoped to bootstrap/materialization and did not affect the model-core test/typecheck proof.