Merge origin/main (v1.64.1.0) into garrytan/time-attack-fork-review

The code-smell wave refactored the tunnel start into the staged
startTunnel helper; the merge adopts that structure and threads this
branch's truthful consent strings through its consent parameter (the
receipt now names the isPairAgentEnabled gate at both call sites). The
eng-review PTY warmup takes both improvements: main's resolveClaudeBinary
fallback and this branch's resolveEvalModel kind. Resolver imports
union; carve budgets take the larger of both waves' measured values
(parity suite green on the merged tree); conflicted generated SKILL.md
files regenerated from resolved sources. VERSION stays 1.65.0.0 over
main's 1.64.1.0; CHANGELOG stacks 1.65.0.0 > 1.64.1.0 > 1.64.0.0.
This commit is contained in:
Garry Tan
2026-08-15 10:01:18 -07:00
187 changed files with 2744 additions and 22778 deletions
+3 -3
View File
@@ -11,7 +11,8 @@
//
// See setup-gbrain/SKILL.md.tmpl Step 4 (Path 4) for the contract under test.
import { describe, test, expect } from 'bun:test';
import { test, expect } from 'bun:test';
import { describeE2ETier } from './helpers/e2e-gate';
import * as fs from 'fs';
import * as os from 'os';
import * as path from 'path';
@@ -22,8 +23,7 @@ import { runAgentSdkTest, passThroughNonAskUserQuestion, resolveClaudeBinary } f
// non-deterministic (it sometimes skips Step 8 CLAUDE.md write, sometimes
// shortcuts past the verify helper). The deterministic gate coverage for
// Path 4 lives in test/setup-gbrain-path4-structure.test.ts (free, <200ms).
const shouldRun = !!process.env.EVALS && process.env.EVALS_TIER === 'periodic';
const describeE2E = shouldRun ? describe : describe.skip;
const describeE2E = describeE2ETier('periodic');
// Spin up a stub MCP server that responds to initialize + tools/list.
function startStubMcpServer(opts: { failWithStatus?: number; failBody?: string } = {}): Promise<{ url: string; close: () => Promise<void> }> {