[juliusmarminge]

Problem

When a locked SSH agent cannot sign, each desktop reconnect opens an SSH password prompt even though unlocking the agent is needed. A local OpenSSH integration fixture reproduces this on both password-enabled and key-only hosts: two connection attempts produce two password requests.

Change

Return the original SSH error when OpenSSH reports an agent signing failure, before entering password fallback. Keep existing password, keyboard-interactive, and encrypted-key passphrase fallback unchanged.

Scope and approval

Fixes #23413

Triaged bug via the automated issue flow. The production change is a four-line signing-error guard in the shared SSH package; no renderer, contract, setting, or reconnect-policy changes.

Verification

• vp test run packages/ssh/src/auth.test.ts packages/ssh/src/tunnel.test.ts: 16 passed at fa4f43f61f. The same final test fixture against base production has 3 failures and 13 passes, specifically the signing-error prompt assertions: red output, green output.
• vp run check:changed passes: formatting, lint, unused-code checks, typechecks for four affected packages, and the two focused test files (16 tests). Typechecks report existing suggestions on unchanged code.
• Linux x86_64, OpenSSH 9.6p1: real SshEnvironmentManager, real SSH client/local sshd, and a protocol-agent fixture advertising an accepted ED25519 key. Each row runs two connection attempts under identical conditions. After the fix, both failures retain the original SshCommandError, exit 255, and byte-identical stderr. Validated assertions.

Runtime fixture	Before: 4040867bb4	After: fa4f43f61f
Agent refuses signing; password-enabled host	2 prompts	0 prompts; signing error preserved
Agent disconnects during signing; password-enabled host	2 prompts	0 prompts; communication error preserved
Agent refuses signing; key-only host	2 prompts	0 prompts; signing error preserved

The tests also retain prompts for identity-listing failures, ordinary password/keyboard-interactive denials, and standalone publickey denials. A real encrypted-key SSH control confirms that a key-only host can require a passphrase through askpass.

Evidence index and reproducible fixture scripts. The visual matrix is empty because no render paths changed; these are service transcripts rather than desktop captures. Independent proof audit: sufficient for all six operational rows.

Unverified: macOS/Electron IPC and dialog presentation, real 1Password/Secretive lock/unlock and recovery, Windows, and successful manager-mediated passphrase retry. cannot-verify desktop.

Models: Claude Opus 5.5 and GPT-6.1 Sol; harnesses: Claude and Codex in T3 Code.

[macroscopeapp]

Approvability

Verdict: Needs Human Approval

Macroscope's review found this PR requires human approval — This targeted fix changes the existing SSH authentication fallback by suppressing password prompts after agent signing failures and preserving the original error, with focused regression coverage. Human review is required because the production change is in authentication-sensitive SSH code.

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

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

• Source mode: Auto
• Repository: pingdotgg/t3code
• Branch: fix/issue-23413
• File: .coderabbit.config.ts
• Commit: fa4f43f
• Layers: Repository TS config
• Review profile: CHILL
• Plan: Team
• Run ID: f0e17e82-42c7-49c1-b26a-a3e15909a203

📥 Commits

Reviewing files that changed from the base of the PR and between 96bae54 and fa4f43f.

📒 Files selected for processing: 2

• packages/ssh/src/auth.ts
• packages/ssh/src/tunnel.test.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 1 remain after this review.

📝 Walkthrough

The SSH authentication classifier now excludes agent signing failures from authentication-failure classification. Parameterized tests check the resulting password-prompt behavior alongside agent communication failures and authentication denials.

Changes: SSH authentication

Layer / File(s)	Summary
Agent signing failure classification and prompt tests packages/ssh/src/auth.ts, packages/ssh/src/tunnel.test.ts	isSshAuthFailure excludes normalized agent signing failure messages matching the added pattern. Tests verify environment failure and password-prompt counts for agent errors and authentication denials. No-prompt cases also check the SshCommandError and original stderr.

Priority: ➖ Normal

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Bug fix · Severity of issue fixed: Medium

Suggested reviewers: t3dotgg

Merge Risk: ⚪ Minimal · up to fa4f4

The change appears ready to merge after normal checks; no actionable issue was identified in the supplied evidence.

Pre-merge checks: 4

Check name	Status	Explanation
Description check		The description covers the problem, change, scope, linked issue, focused verification, observed results, and unverified environments. It does not include explicit maintainer approval, but the linked t…
Title check		The title is concise, specific, and accurately describes the main change: agent signing failures no longer trigger SSH password prompts.
Linked Issues check		The PR addresses issue #23413. isSshAuthFailure now excludes OpenSSH sign_and_send_pubkey: signing failed ... from agent: errors. The tunnel therefore returns the original SshCommandError and do…
Out of Scope Changes check		The reported changes are limited to the shared SSH authentication guard and focused tunnel tests. The tests verify the issue behavior and protect existing fallback behavior. No unrelated renderer, con…

Full details: Description check

Explanation

The description covers the problem, change, scope, linked issue, focused verification, observed results, and unverified environments. It does not include explicit maintainer approval, but the linked triaged issue and detailed scope explanation provide sufficient context for this focused bug fix.

Full details: Linked Issues check

Explanation

The PR addresses issue #23413. isSshAuthFailure now excludes OpenSSH sign_and_send_pubkey: signing failed ... from agent: errors. The tunnel therefore returns the original SshCommandError and does not request a password. Added tests cover signing failures, communication failures, key-only hosts, and retained prompts for identity-listing, publickey, password, and keyboard-interactive failures. The change leaves normal agent success and existing fallback paths unchanged.

Full details: Out of Scope Changes check

Explanation

The reported changes are limited to the shared SSH authentication guard and focused tunnel tests. The tests verify the issue behavior and protect existing fallback behavior. No unrelated renderer, contract, setting, reconnect-policy, or production subsystem changes are reported.

✨ Finishing Touches

📝 Generate docstrings

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}

[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	5.0 KiB	5.0 KiB	−7 B (−0.1%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.2 KiB	−7 B (−0.6%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	20.9 KiB	20.9 KiB	0 B (0.0%)	29.3 KiB	✅
Codex	Live turn messages	2	2	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	+7 B (+0.1%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.2 KiB	1.2 KiB	+7 B (+0.6%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.2 KiB	21.2 KiB	0 B (0.0%)	29.3 KiB	✅
Claude	Live turn messages	2	2	0 (0.0%)	8	✅

Baseline: 96bae54 · PR result: fa4f43f · 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: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

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

[juliusmarminge]

Closing in favour of #23437, which makes the same change.