Auto-logging middleware for OpenClaw

[dhawalc]

Goal: Fully automatic logging without manual log_now() calls.

Current state:

• Manual: User must call log_achievement(), log_error(), etc.
• ULTRON has helpers in auto_log.py but still requires explicit calls

Proposed:

• OpenClaw middleware that intercepts all agent messages
• AI-decides what's important (achievements, errors, decisions)
• Auto-logs to AOMS without user intervention

Implementation:

• Hook into OpenClaw message pipeline
• Use local LLM (Ollama qwen2.5:7b) to classify message importance
• Auto-extract: title, outcome, tags
• Silent write to AOMS (fire-and-forget)

Success criteria:

• Zero manual logging needed
• Agent naturally accumulates episodic memory
• Works across all OpenClaw agents

Priority: Medium (nice-to-have, manual logging works for now)