v1.62.0.1 fix: recognize Claude Keychain authentication

[shawnacalia]

Summary

• Make the Claude outside-voice skill query claude auth status --text before applying legacy authentication heuristics, so macOS Keychain-backed Claude Code sessions are recognized.
• Preserve credential-file and ANTHROPIC_API_KEY fallbacks for older CLI versions and API-key-based setups.
• Add generated-skill regression assertions for Codex and all nine supported external hosts.
• Release the fix as 1.62.0.1 with a changelog entry.

Test Coverage

All new behavior is covered by generated-skill assertions. The new CLI probe is pinned in the Codex output and every external-host rendering; the explanatory Keychain guidance is also pinned in the Codex output.

Pre-Landing Review

No issues found. The auth probe is silent, delegates secure-store detection to Claude Code itself, and falls back safely when older clients do not support auth status.

Design Review

No frontend files changed — design review skipped.

Eval Results

• bun test: passed the full free suite.
• bun test test/gen-skill-docs.test.ts: 405 passed, 0 failed.
• bun run gen:skill-docs --host codex --dry-run: all generated Codex skills fresh.
• bun run test:evals: 48 passed, 341 skipped, 12 failed. The failures were environment-limited and unrelated to this diff: Claude-backed routing runs returned no model output/tool calls, and an always-on Swift fixture could not import XCTest.
• bun run build: generated all host skills and compiled binaries, then stopped in the packaging phase because the local shell advertises an unavailable C.UTF-8 locale to Perl. No tracked build artifacts changed.

Scope Drift

Scope Check: CLEAN. The implementation is confined to the Claude skill template, its generator regressions, and required release metadata.

Plan Completion

No plan file detected.

TODOS

No TODO items completed in this PR.

Documentation

Updated CHANGELOG.md for 1.62.0.1. No additional user documentation changes are needed because the behavior and compatibility fallbacks are documented directly in the generated skill instructions.

Test plan

• Generated-skill regression suite passes (405 tests, 0 failures)
• Codex generated-skill freshness check passes
• Public diff and PR text pass gstack-redact with zero findings
• Paid/model-backed eval suite passes in this environment (blocked by unavailable Claude test output)
• Full packaging build passes in this environment (blocked by local Perl locale configuration)

Generated with OpenAI Codex.

[trunk-io]

Merging to main in this repository is managed by Trunk.

• To merge this pull request, check the box to the left or comment /trunk merge below.

After your PR is submitted to the merge queue, this comment will be automatically updated with its status. If the PR fails, failure details will also be posted here

[garrytan]

Closing: this was absorbed or superseded on main by v1.67.0.0 or earlier (see the v1.67 tracker-wave receipts in PR #2604). Thank you!