Fix flaky Cursor/OpenCode E2E behavior and transcript prep timing

[pfleidi]

Summary

• Re-resolve Cursor transcript paths before waiting and increase transcript preparation wait time to avoid missing nested session file writes.
• Reduce Cursor interactive test flakiness by broadening ready-state prompt matching and slightly increasing startup/trust-dialog wait windows.
• Make OpenCode honor per-prompt timeout overrides and raise TestMultiSessionSequential prompt timeouts to reduce timeout-bound failures.
• Harden a checkpoint path-normalization test on macOS by resolving temp dir symlinks before repo initialization.

Validation

• go build ./...
• go vet ./...
• go test -c -tags e2e ./e2e/tests
• mise run test:e2e:canary
• mise run test:e2e --agent cursor-cli "TestInteractive(AttributionMultiCommitSameSession|AttributionOnAgentCommit|ContentOverlapRevertNewFile|MultiStep|ShadowBranchCleanedAfterAgentCommit)$"
• mise run test:e2e --agent opencode TestMultiSessionSequential

---

Note

Low Risk
Low risk: changes are limited to transcript polling behavior and test/e2e timeouts/prompt matching, with minimal impact on core business logic. Main risk is masking genuine failures by increasing waits or broadening readiness regexes.

Overview
Hardens Cursor transcript preparation by extending the wait window and making polling respect ctx deadlines/cancellation using a timer-based sleep.

Stabilizes E2E agent runs by broadening Cursor CLI ready-state prompt matching, increasing startup/trust-dialog wait timeouts, and making OpenCode honor WithPromptTimeout over the E2E_TIMEOUT env var.

Reduces test flakiness across platforms by resolving macOS temp-dir symlinks in a checkpoint path-normalization test, and by raising timeouts in TestMultiSessionSequential via per-prompt overrides.

^{Reviewed by Cursor Bugbot for commit 957579f. Configure here.}

[Copilot]

Pull request overview

This PR reduces E2E flakiness for Cursor and OpenCode agents and improves transcript preparation reliability in the manual-commit strategy, aiming to avoid missed transcript writes and timeout-bound failures.

Changes:

• Adds per-prompt timeout overrides in TestMultiSessionSequential and makes OpenCode read those overrides.
• Broadens Cursor CLI “ready” prompt matching and increases startup/trust wait windows.
• Re-resolves transcript paths before transcript preparation waits to handle Cursor’s flat→nested relocation behavior; hardens a macOS path normalization test via symlink resolution.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
e2e/tests/multi_session_test.go	Passes per-prompt timeout overrides to reduce sequential multi-prompt failures.
e2e/agents/opencode.go	Applies per-prompt timeout configuration when running OpenCode prompts.
e2e/agents/cursor_cli.go	Expands prompt readiness regex and increases wait windows to reduce Cursor CLI flakiness.
cmd/entire/cli/strategy/common.go	Re-resolves transcript paths before preparation waits to avoid polling stale locations.
cmd/entire/cli/checkpoint/checkpoint_test.go	Stabilizes macOS path normalization test by resolving temp dir symlinks.
cmd/entire/cli/agent/cursor/cursor.go	Increases Cursor transcript preparation wait window.

Comments suppressed due to low confidence (1)

e2e/agents/opencode.go:105

• In openCodeAgent.RunPrompt, the per-prompt timeout option (cfg.PromptTimeout) is overridden by the E2E_TIMEOUT env var unconditionally. That means callers (like tests) cannot actually override the timeout when E2E_TIMEOUT is set, which conflicts with the intended “per-prompt override” behavior. Consider making the precedence explicit (e.g., cfg.PromptTimeout > env > default, or at least document that env always wins).

	timeout := a.timeout
	if cfg.PromptTimeout > 0 {
		timeout = cfg.PromptTimeout
	}
	if envTimeout := os.Getenv("E2E_TIMEOUT"); envTimeout != "" {
		if parsed, err := time.ParseDuration(envTimeout); err == nil {
			timeout = parsed
		}
	}

[pfleidi]

Bugbot run

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 957579f. Configure here.}

[computermode]

🙏🏻 🙏🏻 🙏🏻 🙏🏻 🙏🏻 🙏🏻