ec-session-cleaner_skill

Overview

This skill converts raw OpenClaw / Clawdbot session JSONL transcripts into clean, readable Markdown. It removes tool calls, system metadata, and operational noise while preserving the human and assistant conversation. The output is suitable for archives, documentation, and review.

How this skill works

The cleaner parses .jsonl session files, identifies message roles and content, and strips out non-conversational entries such as tool invocations and internal metadata. It reconstructs the dialogue into concise Markdown with speaker labels and preserves timestamps only when relevant. Batch scripts enable processing multiple agent sessions and remote cleanup via SSH for archived datasets.

When to use it

• Preparing conversation archives for human review or publication
• Removing system noise before importing transcripts into search or analysis tools
• Batch-cleaning multiple agent sessions in a data archive
• Converting noisy logs into readable documentation or training examples
• Cleaning remote session files on a server via SSH

Best practices

• Keep original JSONL files unchanged and run the cleaner to produce separate Markdown outputs
• Use batch scripts for large archives to maintain consistent formatting across files
• Verify a sample of cleaned outputs before bulk processing to confirm noise-stripping rules
• Retain a mapping of original file names to cleaned outputs for traceability
• Run the cleaner in a controlled environment when accessing remote sessions over SSH

Example use cases

• Archive maintenance: convert years of agent sessions into human-readable transcripts for storage.
• Knowledge base creation: extract clean dialogues to populate documentation or FAQ examples.
• Audit and compliance: produce sanitized conversation records without internal tool traces.
• Research: prepare corpora of assistant-user exchanges for qualitative analysis.
• Remote cleanup: run scripts to fetch and clean sessions stored on a remote host via SSH.