Add suite selection to agentlab run

[Jordak]

Previously, agentlab run accepted exactly one --task, so running a suite required external shell loops. That pushed task discovery, ordering, per-task status, and aggregate failure reporting outside the CLI.

That shape made repeated suite pilots more error-prone because users had to recreate orchestration details around the evaluation harness each time.

This PR adds --suite <path> as a mutually exclusive alternative to --task. Suite runs discover task bundles with the existing task discovery helpers, execute discovered tasks sequentially in deterministic order, reuse the same trial configuration for each task, print per-task summaries, and return a nonzero aggregate status when any suite task fails. It also adds --dry-run so users can resolve a task or suite and inspect what would run without starting trials.

GitHub issue #120 remains open for integration-owner landing and closure.

Readiness evidence: #120
Readiness verdict: Ready to Implement

Validation:

• python3 -m unittest tests.test_cli.CliOutputTest.test_run_parser_accepts_jobs tests.test_cli.CliOutputTest.test_run_parser_accepts_suite tests.test_cli.CliOutputTest.test_run_parser_rejects_task_and_suite_together tests.test_cli.CliOutputTest.test_handle_run_preserves_cli_summary_from_trial_execution tests.test_cli.CliOutputTest.test_handle_run_executes_suite_tasks_sequentially tests.test_cli.CliOutputTest.test_handle_run_reports_suite_no_matches tests.test_cli.CliOutputTest.test_handle_run_returns_nonzero_for_suite_trial_failures
• python3 -m unittest tests.test_cli.CliOutputTest.test_run_parser_accepts_dry_run tests.test_cli.CliOutputTest.test_handle_run_dry_run_resolves_task_without_executing_trials tests.test_cli.CliOutputTest.test_handle_run_dry_run_resolves_suite_without_executing_trials
• python3 -m unittest tests.test_cli
• python3 -m unittest tests.test_tasks tests.test_trial_execution
• python3 -m unittest tests.test_cli tests.test_tasks tests.test_trial_execution
• python3 -m agentlab run --agent codex --suite tasks/starter --dry-run --trials 1 --jobs 1
• git diff --check
• pre-commit task bundle integrity check during commits

[Jordak]

Agent Review

Scope: PR #121, fc257379..1d777821
Pass: 1 fresh
Review packet: clean panel summarized in /private/tmp/issue120-review-loop-report.html
Reviewers: P1-R1 correctness, P1-R2 tests-regressions, P1-R3 structural-depth
Effort metadata: implement-github-issue prescribed high from AgentOS defaults; review-loop prescribed medium but run in this same high-effort thread; review-pass prescribed high from AgentOS defaults; effective model/effort not reported by subagent runtime.

Issue families:
None. All three eligible clean-context reviewers returned No new findings.

Declined issue families:

• None.

Resolved ask-user decisions:

• None.

Unresolved ask-user blockers:

• None.

Validation notes:

• python3 -m unittest tests.test_cli.CliOutputTest.test_run_parser_accepts_jobs tests.test_cli.CliOutputTest.test_run_parser_accepts_suite tests.test_cli.CliOutputTest.test_run_parser_rejects_task_and_suite_together tests.test_cli.CliOutputTest.test_handle_run_preserves_cli_summary_from_trial_execution tests.test_cli.CliOutputTest.test_handle_run_executes_suite_tasks_sequentially tests.test_cli.CliOutputTest.test_handle_run_reports_suite_no_matches tests.test_cli.CliOutputTest.test_handle_run_returns_nonzero_for_suite_trial_failures
• python3 -m unittest tests.test_cli
• python3 -m unittest tests.test_tasks tests.test_trial_execution
• python3 -m unittest tests.test_cli tests.test_tasks tests.test_trial_execution
• git diff --check
• pre-commit task bundle integrity check during commit

Terminal condition: completed. Final fresh clean-context review-pass evidence found no unresolved auto-fix or ask-user blockers. PR #121 is ready for integration-owner review; merge, GitHub issue #120 closure, and branch deletion remain outside this workflow.

[Jordak]

Agent Review

Scope: PR #121, fc257379..7cb8c589
Pass: 2 fresh, after --dry-run follow-up
Review packet: clean panel summarized in chat/review-loop result
Reviewers: P2-R1 correctness, P2-R2 tests-regressions
Effort metadata: review-pass prescribed high from AgentOS defaults; effective model/effort not reported by subagent runtime.

Issue families:
None. Both eligible clean-context reviewers returned No new findings.

Declined issue families:

• None.

Resolved ask-user decisions:

• None.

Unresolved ask-user blockers:

• None.

Validation notes:

• python3 -m unittest tests.test_cli.CliOutputTest.test_run_parser_accepts_dry_run tests.test_cli.CliOutputTest.test_handle_run_dry_run_resolves_task_without_executing_trials tests.test_cli.CliOutputTest.test_handle_run_dry_run_resolves_suite_without_executing_trials
• python3 -m unittest tests.test_cli
• python3 -m agentlab run --agent codex --suite tasks/starter --dry-run --trials 1 --jobs 1
• git diff --check
• pre-commit task bundle integrity check during dry-run follow-up commit

Terminal condition: completed for updated head 7cb8c589. Final fresh clean-context review-pass evidence found no unresolved auto-fix or ask-user blockers. PR #121 remains ready for integration-owner review; merge, GitHub issue #120 closure, and branch deletion remain outside this workflow.