fix(cli): exit non-zero when a resumed task does not complete (#1005 follow-up)

[frankbria]

Post-merge review finding on #1020.

cf work resume printed "Task execution failed" and exited 0, so cf work resume <id> && next_step proceeded as if the run had succeeded. work start exits 1 for anything short of COMPLETED (app.py:2610), and #1020's body promised parity with it.

Matched to work start exactly — BLOCKED and FAILED both exit 1. The reviewer suggested gating on FAILED alone, on the grounds that blocking again is a normal outcome of a resume. That is true, but it is equally true of work start, which exits 1 anyway; one predictable rule across both commands beats two that differ by a case.

Parametrized test pins all three outcomes (COMPLETED→0, FAILED→1, BLOCKED→1).

tests/ui/test_resume_starts_worker_1005.py: 13 passed, ruff clean.

[github-actions]

Claude finished @frankbria's task in 3m 32s —— View job

---

Precision bug review — complete

Hunted for concrete defects only (logic errors, null/error paths, security, data loss, race conditions, breaking changes, resource leaks).

• Gather PR diff + existing comments
• Read app.py resume + work start context
• Read the test file
• Confirm/verify findings against real code
• Post summary

✅ No concrete defects found. Verdict + reasoning posted as a summary comment below.

Scope note: the BLOCKED→exit-1 design choice (vs. FAILED-only) is a deliberate parity decision with work start and was explicitly defended in the PR body — that's an architecture/design call, out of scope for this defect-only review.

[github-actions]

✅ GLM precision review: no defects found.

Reviewed the production hunk in codeframe/cli/app.py (the new raise typer.Exit(1) for BLOCKED/FAILED) against its work start counterpart and the surrounding control flow:

• The gate state.status in (AgentStatus.BLOCKED, AgentStatus.FAILED) is byte-for-byte identical to work start at app.py:2610 — no inverted/wrong condition.
• state is guaranteed non-None here (already dereferenced at line 2718), so no None-deref on the new line.
• The raise typer.Exit(1) sits after the inner try/except Exception and inside the outer try, whose handlers are FileNotFoundError/ValueError — typer.Exit is neither, so it propagates to the shell correctly (no silent swallow).
• --no-execute branch is untouched (still exits 0); the existing status=None recorder sentinel also falls through to 0, so no regression to test_cli_resume_executes.
• Test helper's getattr(AgentStatus, status) is valid for all three parametrized members.

View job run