⚙

Mission ControlOperator

Connected

pieter_pan

Systemarchitektur

Agents, Memory-Layers, Defense-Crons, Rules — live aggregiert vom Homeserver.

Architecture aktiv0/13 agents · 56 cronsArchitecture2026-05-13T13:30:01.649953+00:00

Section links

➕Overview 🏗️Architecture 📊Analytics 💰Costs 📈Trends 📂Files 📅Calendar

System Health

Atlas Session

Crons

Graph Edges

1,900

Active Agents

0/13

SSE: connectinglast update 2026-05-13T13:30:01.649953+00:00

🤖 Agents (13)

Status from latest session activity (active = within 5 min)

codexidle

0 sessions

defaultidle

0 sessions

efficiency-auditormonitoring

443 sessions · 2 KB

frontend-gurumonitoring

418 sessions · 2 KB

hermesmonitoring

363 sessions · 2 KB

jamesidle

3 sessions · 10 KB

mainmonitoring

869 sessions · 2 KB

memory-pipelineidle

0 sessions

sparkmonitoring

398 sessions · 2 KB

sre-expertmonitoring

2044 sessions · 2 KB

system-botmonitoring

609 sessions · 2 KB

test-lockidle

0 sessions

workeridle

0 sessions

🧠 Memory Layers

QMD-Index

indexed via qmd update */30

L1.5

KB-Compiler

nightly

Rules + Facts

61 rules · 513 facts

L2.5

Graph

1900 edges

Retrieval-Feedback

hourly

Budget-Meter

no readings

Dashboard.md

nightly 04:30 UTC

🛡️ Defense-Crons (56)

T1-realtime: 3 · T2-2min: 2 · T3-5min: 6 · T4-10min: 2 · T5-15min: 3 · T6-30min: 4 · T7-hourly: 6 · T8-daily-or-weekly: 30

T1-realtime (3)

* * * * * → openclaw-config-guard.sh
* * * * * → crontab-schema-gate.sh
* * * * * → discord-trigger-poller.py

T2-2min (2)

*/2 * * * * → flock
*/2 * * * * → state-collector.py

T3-5min (6)

*/5 * * * * → cost-alert-dispatcher.py
*/5 * * * * → mc-critical-alert.py
*/5 * * * * → memory-budget-meter.sh
*/5 * * * * → session-size-guard.py
*/5 * * * * → mcp-taskboard-reaper.sh
*/5 * * * * → session-rotation-watchdog.py

T4-10min (2)

*/10 * * * * → atlas-orphan-detect.sh
*/10 * * * * → session-health-monitor.py

T5-15min (3)

*/15 * * * * → self-optimizer.py
*/15 * * * * → trajectory-hygiene.py
*/15 * * * * → cron-runs-tracker.py

T6-30min (4)

23,53 * * * * → session-size-alert.sh
*/30 * * * * → openclaw-memworker-wrap.sh
15,45 * * * * → qmd-native-embed-cron.sh
*/30 * * * * → phase-7-readiness.py

T7-hourly (6)

0 */1 * * * → r48-board-hygiene-cron.sh
40 * * * * → openclaw-memworker-wrap.sh
0 * * * * → mc-ops-monitor.sh
0 * * * * → rules-render.sh
5 * * * * → qmd-pending-monitor.sh
23 * * * * → minions-pr-watch.sh

T8-daily-or-weekly (30)

1-59/5 * * * * → sprint-debrief-watch.sh
5-59/15 * * * * → r49-claim-validator.py
45 2 * * * → openclaw-memworker-wrap.sh
0 5 * * 0 → memory-orchestrator.py
0 4 1 */3 * → memory-orchestrator.py
7 */6 * * * → memory-size-guard.sh
0 */6 * * * → script-integrity-check.sh
0 */6 * * * → flock
0 3 * * * → cleanup.sh
0 3 * * * → config-snapshot-to-vault.sh
0 3 * * 0 → build-artifact-cleanup.sh
10-59/30 * * * * → cron-health-audit.sh
0 */6 * * * → alert-dispatcher.sh
15-59/30 * * * * → session-janitor.py
2-59/5 * * * * → cpu-runaway-guard.sh
0 6 * * * → agents-md-size-check.sh
3-59/5 * * * * → session-size-guard.py
4-59/5 * * * * → mcp-qmd-reaper.sh
0 8 * * * → vault-search-daily-checkpoint.sh
1-59/5 * * * * → per-tool-byte-meter.py
20-59/30 * * * * → architecture-snapshot-generator.py
2-59/5 * * * * → arch-deploy-readiness-check.sh
05 21 * * * → daily-ops-digest.py
10-59/15 * * * * → billing-alert-watch.sh
30 */6 * * * → vault-frontmatter-validator.py
0 5 * * * → memory-schema-gate.py
30 5 * * * → run-memory-tests-with-alert.sh
0 2 * * * → backup-memory.sh
0 */6 * * * → rem-quality-sample.py
0 4 * * * → bak-retention.sh

📜 Rules (59)

Agent-Targeting-Regeln (2)

R11Runtime-ID vs Alias nicht verwechseln [active]
R12Worker-Agents dürfen kein LTM schreiben [active]

API-Regeln (3)

R1Verify-After-Write ist Pflicht [active]
R2Kein unreplaced `{placeholder}` in Task-Description [active]
R3Atlas meldet keinen Erfolg ohne GET-Verify [active]

Atlas-Governance (4)

R49Atlas Anti-Hallucination Claim-Verify-Before-Report [active]
R54MCP-Not-Connected erst als Session-/Gateway-Korrelation triagieren [active]
R55Gateway-Restart heilt keine stale MCP-Session-Runtimes [active]
R57Atlas terminal results use canonical Stage-7 format [active]

Board-Hygiene (1)

R48Board-Hygiene-Cron auto-cancel stale drafts [kandidiert]

Build / Deploy-Regeln (2)

R7Kanonische Build-Sequenz (nicht `deploy.sh`) [active]
R8Jeder Edit bekommt `.bak-<scope>-<datum>` [active]

Build & Code-Safety (3)

R26Server-Only Import-Disziplin [resolved]
R27Legacy-Task nach Root-Cause-Fix [resolved]
R28Operator-Lock-Respekt (geplant, Phase 2 Stabilization-Plan) [active]

Build-Deploy-Regeln (17)

R29Build-Storm-Debounce [active]
R30MCP-Taskboard-Server-Zombies [pending]
R31API-Ghost-State (List vs Get divergiert) [pending]
R32Dispatch-Gate Atlas-Sonderfall [pending]
R33Cron-Script-Pfad-Integrität [active]
R34Bootstrap-Limit für MEMORY.md (Agent-Bootstrap-Truncation) [resolved]
R35Atlas-Self-Report ≠ Board-Truth [active]
R36Agent-Session-File-Size-Creep [pending]
R37Atlas-Orchestrator-Tasks nicht via Auto-Pickup [resolved]
R38MCP-Zombie-Defense-in-depth (existierender Reaper + Alert) [resolved]
R39Atlas-main braucht Session-Resume-Pattern [active]
R40Stall-Detection-Thresholds sind Kern-Infra [resolved]
R41Memory-Retrieval: QMD vor File-Read [active]
R46Parallel-Deploy-Serialization [active]
R50Session-Lock-Governance fuer Auto-Pickup [active]
R42Deploy-Restart-Discipline via mc-restart-safe [active]
R52Auto-Pickup Silent-Fail-Detection [active]

Codex-Governance (1)

R56Vault-SSOT und Sprint-Read-Order respektieren [active]

Config-Regeln (4)

R4openclaw.json NIE direkt editieren [active]
R5Kanonischer MC-Service ist User-Level, Port 3000 [active]
R6`worker-pickup-loop.py` bleibt tot [active]
R51Schema-Validation-Gate fuer openclaw.json [active]

Governance (1)

R47Scope-Lock-auf-Plan-Doc nicht Task-ID [active]

Hygiene (1)

R53Config/Scripts Daily Snapshot in Vault [active]

Integrations-Regeln (2)

R9Discord-Webhook-Calls brauchen User-Agent [resolved]
R10Alerts laufen NICHT über MC-API [active]

Multi-Agent-Koordinations-Regeln (7)

R18mc-ops-monitor ist read-only-alerting [active]
R19heartbeat darf keinen Subagent spawnen für terminale Tasks [active]
R20worker-monitor ist run-lifecycle-only [active]
R21Layer-Cleanup-Tasks brauchen Script-Referenz-Check [active]
R22Task ohne Result-Receipt ≠ erfolgsfrei [resolved]
R23Retry-Task nur bei Parent in failed/error-State [active]
R58Phased Board Sprint Gates muessen maschinenlesbar persistiert sein [active]

Multi-Agent-Orchestration (2)

R45Sub-Agent-Receipt-Discipline [active]
R44Board-Discipline: Board-Task required before sessions_spawn [active]

Naming & Runtime-Regeln (2)

R24Runtime-ID vs Display-Alias Disziplin (verschärft) [resolved]
R25workerLabel muss beim Dispatch gesetzt werden [resolved]

Operator-Eingriff-Regeln (5)

R13Operator greift bei Instabilität ein, nicht bei langsamkeit [active]
R14Tangenten → `spawn_task`, nie mitnehmen [active]
R15Deploy-Sequenz ist atomar im Agent-Turn [active]
R16V8-Heap-Limit muss explizit sein [active]
R17systemd MemoryMax > V8 Heap-Limit [active]

Receipt-Discipline (1)

R60Sprint-Outcome v1.1 Pflichtfelder [active]

Routing-Governance (1)

R59Hermes scoped-build Forge/SRE sprints must dispatch to sre-expert, not Atlas/main [active]

🔗 Data Flows

Static topology (architecture manifest, not filesystem-derived)

vault → L1 indexing(*/30)
vault.10-KB → L1.5 compilation(nightly)
facts → L2.5 edge-inference(hourly)
rules → L2.5 edge-inference(hourly)
L2.5 → L3 feedback(hourly)
agent.main → L5 size-monitor(*/5)
L5 → preflight.gate1 gateblocking
agent.main → vault writes
L1 → L6 aggregation(nightly)
L2 → L6 aggregation(nightly)
L2.5 → L6 aggregation(nightly)
L3 → L6 aggregation(nightly)
L5 → L6 aggregation(nightly)

🏃 Active Sprint

unknown · started 2026-04-28T18:32:16Z

source: /home/piet/vault/_agents/_coordination/live/2026-04-28_2032_codex-desktop_terminal-codex-coordination.md

🔌 MCP Servers (2)

qmdrunning

6 processes

taskboard-mcprunning

2 processes