| name | session-logs |
| description | Use when the user asks about earlier conversations, prior outputs, historical context, or past work that may live in OpenClaw memories, QMD stores, or session logs. |
| user-invocable | true |
| disable-model-invocation | false |
Session Logs
Use this skill to recover prior context from local OpenClaw artifacts.
Search order
/workspace/MEMORY.md/workspace/memory/{{ PRIVATE_DATA_DIR }}/sessions/~/.openclaw/agents/*/qmd/sessions/~/.openclaw/agents/*/sessions/
When to use
- "What did we say before?"
- "Find the previous discussion about X"
- "Search older sessions"
- "What did OpenClaw decide last time?"
Tools
- prefer
rg for filename and content filtering
- use
jq for structured extraction from JSONL transcripts
Useful patterns
Search memories first:
rg -n "phrase" /workspace/MEMORY.md /workspace/memory
Search indexed and raw sessions:
rg -l "phrase" {{ PRIVATE_DATA_DIR }}/sessions ~/.openclaw/agents/*/qmd/sessions ~/.openclaw/agents/*/sessions 2>/dev/null
Extract assistant text from JSONL:
jq -r 'select(.type == "response_item" and .payload.type == "message" and .payload.role == "assistant") | .payload.content[]? | select(.type == "output_text" or .type == "input_text") | .text' <session>.jsonl
Rules
- Prefer memory notes before raw transcripts.
- Quote only the lines needed.
- Include the source path when it helps verification.
- If memory notes and raw session logs disagree, say so.
