feat(recommendation): interactive TUI for get and list

[jariy17]

What

Adds an interactive TUI for the eval recommendation get and list commands, mirroring how batch-evaluation does it.

• list/screen.tsx — RecommendationListScreen: paginated table picker (name / type / status / updated UTC), selecting a row navigates to the get view.
• get/screen.tsx — RecommendationGetJsonScreen: raw-JSON detail view (no metadata hub, matching batch-evaluation).
• screen.tsx — RecommendationScreen: parent router menu.
• Router now opens the TUI on a bare recommendation invocation (withTuiOnEmptyFlagsAndArgs + renderTui) and marks get/list as TUI-supported; start/delete stay command-line-only.
• eval menu now lists recommendation as a TUI group; routes wired in Root.tsx.

Notes

• Used PaginatedTablePicker directly in the list screen rather than extracting a RecommendationPicker component — list is the only picker consumer (unlike BatchEvaluationPicker, reused by the start flows).
• Updated CliOnlyScreen.test.tsx: the eval menu no longer lists recommendation as CLI-only, and the "all subcommands CLI-only" example now points at ondemand.

Test

• bun run typecheck clean
• bun test src/handlers/eval src/router src/components → 671 pass, 0 fail

[agentcore-devx-automation]

Claude Security Review: no high-confidence findings. (run)

[agentcore-devx-automation]

AgentCore Harness Review

Verdict: Looks good

Clean, focused change that mirrors the existing batch-evaluation TUI pattern one-for-one. I checked the parts most likely to drift:

• Routes in Root.tsx (parent, list, bare-get redirect, get/:id) match the batch-evaluation shape.
• Router wiring in handlers/eval/recommendation/index.tsx (withTuiOnEmptyFlagsAndArgs + renderTui + supportedTuiCommands("get", "list")) matches batch-evaluation.
• listRecommendations(token, pageSize, undefined, opts) argument order matches Core.listRecommendations(nextToken, maxResults, statusFilter, options).
• RecommendationRow/toRow fields (name, type, status, updatedAt) all exist on the SDK RecommendationSummary (confirmed against the get golden fixture).
• The CliOnlyScreen.test.tsx updates are factually correct: ondemand really does only expose evaluate and simulate, and recommendation correctly drops off the eval menu's CLI-only list now that it has a screen.
• eval/index.tsx adds recommendation to supportedTuiCommands consistent with the new screen.

No excessive mocking; screens use real query-client behavior around core.eval.*. No telemetry gap unique to this PR — the sibling batch-evaluation screens don't instrument either.

Non-blocking observation: batch-evaluation has a companion batch-evaluation.screen.test.tsx covering the picker/menu/get UI, but no equivalent recommendation.screen.test.tsx was added. The compiled-tree walk in CliOnlyScreen.test.tsx gives some indirect coverage, and the fixture test covers the CLI path, so this is a suggestion rather than a required change — worth adding in a follow-up for parity.

Good to merge.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.05%. Comparing base (d9ab959) to head (f213da3).
⚠️ Report is 11 commits behind head on refactor.

Additional details and impacted files

@@            Coverage Diff             @@
##           refactor    #2279    +/-   ##
==========================================
  Coverage     97.04%   97.05%            
==========================================
  Files           566      569     +3     
  Lines         39409    39511   +102     
==========================================
+ Hits          38244    38346   +102     
  Misses         1165     1165            

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[agentcore-devx-automation]

Claude Security Review: no high-confidence findings. (run)