Daily Logs

Type: Layer 3 of OpenClaw's 5-layer memory system

Definition

Daily logs are append-only files at memory/YYYY-MM-DD.md that record operational events each day. They are layer 3 of the 5-layer memory system. Entries include: projects initiated and completed, failures and resolutions, quality gate passes and fails, operator decisions and their rationale, notable findings, and any system changes or discoveries. They form a chronological record that survives session restarts and provides continuity across gaps in work.

How It Works

At the end of each working session (or throughout the day), the agent appends to the daily log file. The naming convention is memory/YYYY-MM-DD.md — the date of the work session. Files are never overwritten; they're only appended to. This preserves the complete history of work done.

Entries follow a consistent format: timestamp (optional), event type, description. Examples: "## 2026-04-27", "### Project initiated", "### Gate failed: output file empty, sent back for revision", "### Operator decision: use Postgres over MongoDB for persistence layer, citing better JSON support".

Daily logs serve two purposes: (1) they provide continuity — if work resumes two weeks later, the log shows what was done, what failed, what the operator decided, and why; (2) they feed the weekly memory maintenance process — a cron job reads all logs from the past week, identifies patterns and learnings, and updates MEMORY.md accordingly.

The logs are especially valuable in multi-agent work. When a sub-agent completes, it writes its results to the log. When the Router validates a gate, it records the outcome. This creates an audit trail that survives the sub-agent session's termination.

Key Properties

  • Append-only — never overwrite; entries only added to preserve history
  • Date-based naming — memory/YYYY-MM-DD.md format; one file per day
  • Layer 3 of 5-layer system — sits between MEMORY.md (persistent facts) and project context (per-project state)
  • Content types — project initiations, completions, failures, resolutions, gate outcomes, operator decisions, findings
  • Multi-agent audit trail — sub-agent results and gate validations recorded for later reference
  • Weekly maintenance input — cron job reads past week's logs to update MEMORY.md with patterns and learnings
  • five-layer-memory — daily logs are layer 3 of the 5-layer memory system
  • context-window — daily logs persist when session context would be compacted
  • ralph-protocol — failure events recorded in daily logs as part of audit trail

Source Chapters