reflect-scheduled-checks

Automating scheduled checks involves executing a single due interval-triggered check per invocation and routing outcomes through reflection-judge and proposal-triage. This approach gates findings for daily maintenance routines while preserving continuity across sessions via state persistence.

State management for interval checks relies on persisting data in state/reflection-state.json under the scheduled_checks key. This mechanism preserves continuity across sessions by recording outcome categories and writing a Progress Log entry to the session file.

Logging progress for scheduled checks appends a Progress Log entry to .claude-code-hermit/sessions/SHELL.md. This records traceability for operator review by capturing outcome categories like actionable, contextual, empty, unavailable, error, and skipped.

Scheduled checks process only one due interval per invocation to avoid cascading results. This design gates findings through the proposal pipeline sequentially, ensuring consistent state updates and preventing overwhelming downstream routing.

Scheduled check automation applies specifically to codebases that register scheduled checks. Without registered intervals, the invocation finds no due checks to execute and cannot route findings through the proposal-triage pipeline.

Scheduled check findings support six outcome categories: actionable, contextual, empty, unavailable, error, and skipped. These categories gate results through the proposal-triage process to determine if findings require operator action or contextual review.