fix(ssh): report remote package installation failures accurately

[juliusmarminge]

Related #10084.

When a remote package installation failed with ETARGET, the SSH runner discarded its exit status and claimed that installation succeeded but needed a C compiler. A failed installer that printed an executable path could even proceed to launch it.

Check the installer's exit status before accepting its stdout. Keep npm's original stderr and report installation failure without guessing its cause. The existing zero-exit/no-executable compiler advice remains intact, as do package versions, installer arguments, retries, timeouts, Node overrides and direct CLI execution.

Verification

• Before the fix, six actual shell-process controls failed and ten passed. Afterward, all sixteen pass. The complete focused runner, tunnel and command run passes all 36 tests; an independent orchestrator run also passed all 36.
• Both npx and npm exec paths cover ETARGET, another nonzero error, failure with a valid path on stdout, zero-exit/no-executable, successful install, an existing CLI and a Node script override. Existing PID, graceful shutdown and restart controls pass.
• SSH package typechecking, targeted formatting and lint pass.

The tests execute the production-generated script through real /bin/sh, using owned synthetic installer executables and no registry access. They reproduce the diagnostic boundary, not the original package-publication timing, remote Node22/npm11 environment, SSH desktop dialog or successful reconnect. No version fallback or publication change is included. Screenshots do not apply to this shell-only fix. All current-head executed CI, Macroscope correctness, Effect Service Conventions and approvability checks pass. No unresolved review threads remain. Cursor review is paused at the team spending limit; CodeRabbit automatic reviews are disabled. Final merge review is pending.

Prepared by GPT 6 Astra via Codex in T3 Code.

Note

Fix require_installed_t3_cli to report package-manager failures in SSH runner

• The require_installed_t3_cli shell helper in tunnel.ts now checks the installer command's exit status. When the package-manager command fails, it prints an installation failure message with the package spec and exits with status 1 instead of silently proceeding.
• When the installer succeeds but emits no executable path, the existing native-dependency/toolchain diagnostic is unchanged.
• Adds a live parameterized test suite in runnerProcess.test.ts covering npx and npm across seven scenarios (resolution failure, network failure, empty output, success, failed install with stdout path, existing CLI, node-script override).
• Risk: the helper now exits immediately on installer failure rather than falling through to the missing-executable diagnostic, so any remote environment where the package manager exits non-zero for non-installation reasons will surface a different error message.

^{Macroscope summarized 4e1843e.}

[cursor]

Bugbot is paused — on-demand spend limit reached

Bugbot uses usage-based billing for this team and has hit its on-demand spend limit.

A team admin can raise the spend limit in the Cursor dashboard, or wait for the next billing cycle to continue.

[macroscopeapp]

Approvability

Verdict: Approved at 4e1843e

Macroscope's review found this PR approvable — This is a focused SSH error-handling fix that preserves npm diagnostics and reports remote package-installation failures instead of misclassifying them as missing executables. The production logic is small, existing successful paths remain unchanged, and the relevant failure modes are covered by regression tests.

^{You can add or adjust custom eligibility rules. Learn more.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.6 KiB	13.5 KiB	−35 B (−0.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.0 KiB	7.0 KiB	+3 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.5 KiB	−38 B (−0.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	57.1 KiB	57.0 KiB	−88 B (−0.2%)	66.4 KiB	✅
Codex	Live turn messages	10	8	−2 (−20.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	−3 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.0 KiB	7.0 KiB	+3 B (+0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.5 KiB	−6 B (−0.1%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.8 KiB	57.8 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	8	8	0 (0.0%)	21	✅

Baseline: 2271a27 · PR result: 4e1843e · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 113.8 KiB
• Claude decoded thread snapshot: 114.5 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.