Adds remote_sync_queue.go with persistent queue + 1s→30s exponential
backoff drainer, max-attempt drop with audit record, ctx.Done() clean
shutdown. sync.go rewired to enqueue dispatches via the queue rather
than the legacy minute-ticker schedule.
CORE_AGENT_SYNC_MAX_ATTEMPTS / CORE_SYNC_MAX_ATTEMPTS env vars override
the default 100-attempt cap.
Targeted tests cover happy-path drain, retry/backoff with stub clock,
cancellation, max-attempt exhaustion, and file-backed restart persistence.
go test was blocked by go.work workspace dep resolution in sandbox; CI
catches the real run.
Co-authored-by: Codex <noreply@openai.com>
Closes tasks.lthn.sh/view.php?id=231
Add warnings for silent filesystem write/delete failures in agentic persistence helpers and record two adjacent hardening gaps for follow-up.\n\nCo-Authored-By: Virgil <virgil@lethean.io>
- prep.go TrackWorkspace mirrors into queue + concurrency store groups
(previously only registry); hydrateWorkspaces reaps filesystem ghosts
(dead PID → failed, persisted back to status.json) so cmdStatus and
out-of-process consumers see coherent state (RFC §15.3)
- sync.go queue read/write goes through go-store first per RFC §16.5
("Queue persists across restarts in db.duckdb"), file remains fallback
for graceful degradation
- statestore.go stateStoreGet helper for go-store-first reads
- tests/cli/restart — new CLI test for RFC §15.7 "dispatch → kill →
restart → no ghost agents" dead-PID reap flow
- 4 new statestore tests: queue group mirror, concurrency refresh,
sync queue persistence, fs ghost reap with disk write-back
Co-Authored-By: Virgil <virgil@lethean.io>
- sync.go: syncBackoffSchedule (1s/5s/15s/60s/5min) with per-push Attempts
and NextAttempt honoured on retry (RFC §16.5)
- runSyncFlushLoop: ticks every minute from OnStartup when API key present,
drains the queue without re-scanning workspaces
- SyncPushInput.QueueOnly: lets flush loop drain without triggering a full
workspace scan (prevents duplicate pushes)
- Sync ledger at .core/sync/ledger.json: fingerprints keyed by workspace
name + (updated_at, runs); skips already-synced workspaces until fresh
activity
- docs/RFC-AGENT.md: synced from plans/code/core/agent/RFC.md with latest
AgentPlan status enum, complete capability, pr.close/branch.delete,
indexed_at/org brain fields
Co-Authored-By: Virgil <virgil@lethean.io>
