fix(cron): classify no-agent failures as script errors

[visualfox-ch]

Summary

• classify non-zero script exits in no_agent cron jobs as script failures
• preserve deterministic script-only execution while making failure summaries actionable
• add regression coverage for the scheduler result contract

Test plan

• python -m pytest tests/cron/test_cron_no_agent.py -q -o 'addopts='
• ruff check cron/scheduler.py tests/cron/test_cron_no_agent.py

Supersedes #63178 with the same focused fix rebased onto current main.

[alt-glitch]

This was generated by AI during triage.

Related: rebased resubmission of the still-open #63178 (same focused no_agent script-failure classification fix). Also part of the cron failure-classification family tracked under #59379. Flagging the #63178/#64474 pair so a maintainer can pick the canonical one to merge.

[visualfox-ch]

Update: #63178 is now closed as superseded by this current-main replacement.

[tonydwb]

Code Review Summary

Verdict: Approved

Fix for cron error classification when no-agent failures occur. Classifying these as script errors instead of silent failures. 31 additions, 0 deletions. Clean fix.

---

Reviewed by Hermes Agent

[teknium1]

Thanks for the focused regression fix. Current main still sends every failed job through _summarize_cron_failure_for_delivery() at cron/scheduler.py:3620, while that helper classifies 429, timeout, and auth-like text as provider failures at cron/scheduler.py:62-89. This conflicts with the no-agent execution path, which returns failed script output before any agent/provider work at cron/scheduler.py:2591-2638.

The proposed early no_agent branch is correctly scoped to the classifier, preserves the existing bounded one-line error treatment, and its parametrized cases cover the three provider-like strings that currently misclassify script failures. The current surrounding context remains compatible with the patch despite the branch being behind current main.

Automated hermes-sweeper review.