v13-path.md substrate update: D3 scope additions for backup, FS ingestor, cadence design, MX metrics

Context

learn/agentos/v13-path.md (authored 2026-05-08 by @neo-opus-4-7 as the chief-architect deliverable) defines D3 (Orchestrator Daemon Architecture) scope and M3-M4 milestone exit gates. Operator alignment-check 2026-05-09: confirms the SHAPE is correct but surfaces 4 substantive scope gaps the doc currently misses.

Empirical audit findings (grep -nE "backup|filesystem|cadence|frequency" learn/agentos/v13-path.md):

Prescription

Single substrate-doc PR updating learn/agentos/v13-path.md (no new files; surgical additions only):

§3 D3 Orchestrator Daemon Architecture — Scope: Add 3 bullets after existing 7:

• Daily backup with rotation cap (PRIO 0 — non-negotiable; manual backup precedes all chroma/graph mutation work)
• Filesystem ingestor for graph (recovers graph completeness after files re-enter the workspace)
• Cadence design — graph-blocking awareness, time-windowed scheduling for high-impact tasks

§3 D3 — New subsection D3.1 "Cadence Design":

• Graph-processing 10min blocking constraint on add_memory
• Tradeoff matrix: (a) explicit time-windows (low-activity hours), (b) cooperative session-coordination flag, (c) adaptive frequency-throttling
• Per-task fairness model: backup daily-fixed, summary burst-then-steady, KB-sync 30min steady, dream cycle low-activity-only
• Defers to design ticket orchestrator-cadence-design (file under Phase A umbrella)

§4 M3 Exit Gate — augment: Add: "Backup PRIO 0 task registered + retention sweep implemented + manual backup discipline (#10780) operational"

§4 M4 — augment: Add FS ingestor as scoped sub-task within "Migrate Decomposed Daemon Services to Orchestrator"

§8 Outcome Metrics — augment: Add: "sandman_handoff regression closed — auto-prioritization MX feedback loop restored (was working pre-PR-#10863, regressed at config-default-flip)"

§5 Tickets to File — augment: Link the 3 Phase 0 tickets (orchestrator-daily-backup, gh-workflow snapshot enhancement, this update) + Phase A umbrella epic placeholder

Acceptance Criteria

• AC1 — §3 D3 scope updated with 3 new bullets (backup PRIO 0, FS ingestor, cadence design)
• AC2 — §3 new subsection D3.1 Cadence Design landed with tradeoff matrix + design-ticket reference
• AC3 — §4 M3 exit gate adds backup PRIO 0 line
• AC4 — §4 M4 includes FS ingestor sub-task
• AC5 — §8 metrics add sandman_handoff MX outcome with regression-closure framing
• AC6 — §5 ticket links updated for Phase 0 batch + Phase A umbrella placeholder
• AC7 — Cross-family review (per pull-request §6.1) before merge

Avoided Traps

• ❌ Re-write whole sections — surgical additions only; the doc is recently authored (2026-05-08) and largely correct on architectural shape
• ❌ Per-scope-item PRs — single PR keeps the doc coherent; operator did NOT push for splits ("phases: you can choose")
• ❌ Move D3.1 Cadence Design content INTO the doc rather than reference an external design ticket — the doc is the architectural map, not the design playbook; cadence-design needs its own dedicated workspace

Provenance

• Operator directive 2026-05-09: "re-read learn/agentos/v13-path.md => are we on the same page where we want to go"
• Doc author: @neo-opus-4-7 2026-05-08
• Empirical: 4 grep'd gaps confirmed against operator's stated end-state vision (orchestrator scope expansion + cadence model + MX feedback loop)
• Pairs with Phase 0 tickets (orchestrator-daily-backup, gh-workflow snapshot enhancement)

Self-Identification: @neo-opus-4-7 (Claude Opus 4.7, Claude Code) — chief-architect lane, Phase 0.D of post-quad coordination round.