knowledge-base/raw/other/2026-06-07_subconscious-runner-script.md

type	source_url	retrieved	title	tags
other	https://vmd185580.tailf38284.ts.net/hector-bot/knowledge-base	2026-06-07	Subconscious Agent Evidence Runner Script v2.1	subconscious automation cron evidence

Source

Skript lebt unter /home/node/workspace/scripts/subconscious-runner.sh auf dem bot-hector Container.

Der Subconscious Agent generiert täglich um 20:00 CEST einen Evidence-Run über 11 Quellen und schreibt die Ergebnisse in .subconscious/runs/<run-id>/evidence.txt.

Skript-Inhalt

#!/usr/bin/env bash
# Subconscious Agent — Evidence Runner v2.1
set -euo pipefail
RUN_ID="run-$(date +%Y%m%d-%H%M%S)"
RUN_DIR="/home/node/workspace/.subconscious/runs/$RUN_ID"
mkdir -p "$RUN_DIR"
EVIDENCE="$RUN_DIR/evidence.txt"
echo "=== SUBCONSCIOUS AGENT EVIDENCE RUN ===" > "$EVIDENCE"
echo "Run ID: $RUN_ID" >> "$EVIDENCE"
echo "Timestamp: $(date -Iseconds)" >> "$EVIDENCE"
echo "Timezone: Europe/Berlin" >> "$EVIDENCE"
echo "" >> "$EVIDENCE"

Evidence Sources (11 Sections)

1. System Status — Uptime, Disk
2. OpenClaw Status — Version
3. Memory Files — MEMORY.md exists check + line count
4. Recent Memory — Alle memory/YYYY-MM-DD.md Dateien mit Zeilenzahl
5. Current Tasks — tasks/current.md (checked items)
6. Knowledge Base — KNOWLEDGE.md exists check
7. Subconscious State — latest-summary.json
8. Workspace Git — Branch + last commit
9. Skills — ls /home/node/.openclaw/skills/
10. Plugins — ls /home/node/.openclaw/extensions/
11. plur1bus Memory — healthcheck.sh execution

Pipeline

1. Cron (20:00 CEST) → Evidence Runner (Shell) → evidence.txt
2. Subconscious Agent Session (Gemini 2.5 Flash, isolated) liest evidence.txt
3. Generiert bis zu 10 Ideen → Debate → Hard Synthesis (60% Konvergenz)
4. Outcome: accepted / rejected / shelved / deferred
5. Ergebnis in latest-summary.json

Known Issues

• todayRuns Counter stuck seit ~13.04.2026
• Execution Gap — akzeptierte Ideen haben keinen automatisierten Ausführungspfad