[JacobStephens2]

Summary

 Setup → Claude Model
+  show documented aliases as suggestions
   accept any Model ID
   Enter keeps the current value or Claude's own default
   validate a named Model with the existing test call

Closes #424

Evidence

Seam under test: the complete Harness Setup interaction through harness::settings::ask, using scripted terminal input and an isolated fake Claude executable.

• Before: cargo test --bin thirdshift harness::settings::tests::claude_suggests_aliases_and_checks_a_custom_model_and_effort -- --exact failed: no suggestions were printed (left: []).
  After: cargo test --bin thirdshift harness::settings::tests passed all 27 tests, including suggestions, custom IDs, the exact validation invocation, blank defaults, current-value retention, refusals and interruption.
• cargo check --bin thirdshift and cargo fmt --check passed.
• /usr/bin/time -p scripts/affected-tests main: 50 site tests passed; the full Rust run stopped after 515 passes on timed_out_trials_keep_partial_usage_and_exportable_session_ancestry (model Grader session failed). Nextest skipped 66 tests and left 2,506 unrun after the failure. Total selected-run wall time was 826.57s, including 164s Rust compilation and the browser checks.
• The exact failing test passed unchanged on an isolated rerun: cargo test --test experiments timed_out_trials_keep_partial_usage_and_exportable_session_ancestry -- --exact --nocapture (1 passed, 9.80s test execution). The full local Rust suite is not claimed green.

The suggestion list follows Claude Code's documented model settings, checked 2026-10-25. A source comment records the refresh policy; Claude resolves alias versions. The manual and homepage describe the behavior.

Review

Independent thirdshift-code-review against main: Standards 0, Spec 0. Each reviewer read all five changed files, including the generated HTML paragraph. Unread changed files: none. Generated-output exclusions: none.

Merge Danger

Door: two-way

Blast Radius: Setup

Unaddressed findings

• Standards: none.
• Spec: none.
• Security: none. Guidance-mode review of the Setup suggestion change found no introduced finding: the change adds only a static suggestion line before the unchanged Model prompt, keeps free-text input and the existing direct-exec test call (Command::new(cli).arg("-p").args(model_args(choice)), no shell), and fetches no live catalog. cargo test --bin thirdshift harness::settings::tests passes all 27 tests.

Built with codex codex-cli 0.164.0 · gpt-6.1-sol · xhigh

Security review outcome

No unaddressed introduced findings.