Skip to content
T1KTheOneKit

auto-sync-back-when-enabled

Auto Sync-Back When Enabled — Don’t Ask, Just Spawn

Section titled “Auto Sync-Back When Enabled — Don’t Ask, Just Spawn”

Status: always-loaded behavioral rule. When the auto-sync feature flags are ON, the assistant MUST auto-spawn the background sync-back/issue sub-agent the moment a sync-queue reminder fires — without asking the user for approval.

Rule

Section titled “Rule”

If features.autoLessonSync, features.autoLibrarySync, or features.autoIssueSubmission is true in the resolved t1k-config-*.json (check .claude/t1k-config-core.json), then on seeing ANY of these system-reminders:

  • [t1k:lib-sync-queue] (N pending library changes → skill updates)
  • [t1k:lesson-queue] / [t1k:lesson-queued] (pending skill/rule lesson)
  • [t1k:auto-issue] (pending issue submission)

…the assistant MUST immediately spawn the prescribed background sub-agent (kit-developer / t1k-kit-developer, run_in_background: true) per the reminder’s Action block — and MUST NOT pause to ask the user “want me to sync-back?” / “should I open the PRs?” / “shall I propagate this?”.

Asking for approval to sync-back when these flags are ON is a rule violation. The flags ARE the standing approval. Surface the result (PR/issue URL) after the agent returns — do not gate the spawn on a confirmation.

When you still report instead of ask

Section titled “When you still report instead of ask”
  • After spawning: report one line per result (PR opened: <url> / issue filed: <url>), then continue.
  • Genuinely destructive or ambiguous kit operations (force-push, merge, conflict resolution) still follow kit-pr-workflow-boundary.md (consumer opens PR only; never merges). Auto-spawn covers OPENING the PR/issue, not merging it.

When NOT to auto-spawn

Section titled “When NOT to auto-spawn”
  • The relevant flag is false → fall back to offering (the user opted out of automation).
  • The reminder is a duplicate of an entry already submitted this session (dedup cache / writeback submitted:true).
  • You are inside a forked/sub-agent context at recursion depth ≥ 2 (agent-security-boilerplate.md depth guard) — report the queue entry up instead of spawning.

Why

Section titled “Why”

Real session evidence (2026-06-04/05, Amplify/Feel decoupling cook): autoLessonSync + autoLibrarySync were already true, yet the assistant repeatedly offered (“want me to open those PRs?”) instead of auto-spawning, and the kit propagation PRs never opened until the user explicitly said “yes, open those PRs.” The config intent (automate propagation) was defeated by an unnecessary approval step. User directive: “update the kit to make sure you will auto do the sync-back next time without approving from me.”

How to apply

Section titled “How to apply”
  1. Resolve the flag (.claude/t1k-config-core.json → features.autoLessonSync / autoLibrarySync / autoIssueSubmission).
  2. Flag ON + queue reminder present → spawn the background sub-agent NOW (see error-recovery.md spawn pattern + orchestration-rules.md agent routing).
  3. Report the URL; never ask first.
Section titled “Related”
  • error-recovery.md — background sub-agent spawn pattern (this rule removes the “ask first” step when flags are on).
  • orchestration-rules.md — /t1k:sync-back + /t1k:issue are background-only.
  • kit-pr-workflow-boundary.md — auto-spawn OPENS the PR from a consumer; merging stays with the kit maintainer.
  • update-kits-before-sync-back.md — the spawned agent still runs the freshness pre-flight.
  • telemetry.md — the auto-lesson/library/issue pipelines these flags gate.