A high-performance forensic toolset for auditing OpenClaw session history to identify cost regressions and behavioral anomalies.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install session-logs-forensics
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install session-logs-forensics using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Session Logs Analyzer?
The session-logs skill is a specialized forensic utility designed for operators using Openclaw Skills to maintain high-quality AI agent performance. By leveraging powerful command-line tools like jq and ripgrep, this skill enables deep inspection of append-only JSONL session data to trace conversation history, monitor spending, and debug unexpected tool-call drift.
Whether you are troubleshooting a sudden increase in API costs or auditing how an agent handled complex multi-turn interactions, session-logs provides the necessary visibility into the agent's historical state. It transforms raw log files into actionable insights, ensuring your Openclaw Skills remain efficient and predictable across long-running deployments.
Session Logs Analyzer Use Cases
- Retrieving historical context from parent conversations not present in active memory.
- Identifying the root cause of sudden cost spikes in AI agent operations.
- Auditing tool-call frequency to detect behavior regressions or drift.
- Performing incident forensics after a failed or unexpected agent interaction.
- Generating daily cost and usage summaries for agent management.
How Session Logs Analyzer Works
- The skill identifies the target agent session directory located at ~/.openclaw/agents/
/sessions/. - It parses the sessions.json index to map specific session keys to their corresponding unique session IDs.
- Operators utilize jq filters against individual .jsonl files to extract structured data such as message roles, cost metrics, or content types.
- Ripgrep (rg) is employed to perform high-speed text searches across the entire session history for specific keywords or phrases.
- Custom scripts calculate aggregates, such as total session cost or tool-call volume, to establish performance baselines.
Session Logs Analyzer Setup
To use this skill, ensure you have the required binary dependencies installed on your local environment.
# Install dependencies on macOS
brew install jq ripgrep
# Install dependencies on Ubuntu/Debian
sudo apt-get update && sudo apt-get install jq ripgrep
Once installed, you can navigate to your agent's session directory and begin running the forensic queries provided in the documentation.
Session Logs Analyzer Data Schema & Taxonomy
The skill interacts with append-only .jsonl files. Each line represents a specific entry in the conversation history with the following structure:
| Field | Description |
|---|---|
type |
Categorizes the entry as "session" (metadata) or "message". |
timestamp |
ISO 8601 formatted timestamp of the event. |
message.role |
The sender of the message: "user", "assistant", or "toolResult". |
message.content[] |
Array containing text blocks, thinking processes, or tool call details. |
message.usage.cost |
Numerical value representing the cost of the response. |
Session Logs Analyzer Advanced Features
- Cost Outlier Scanning: Rapidly triage sessions that exceed your typical budget baselines.
- Threshold Anomaly Flagger: Automatically flag sessions with suspicious jumps in tool-call volume or token usage.
- Session Regression Comparison: Compare message mix and logic flow between a "healthy" session and a problematic one.
- Forensic Snapshots: Generate compact, shareable summaries of session performance for incident reports.
- Tool Usage Breakdown: Analyze which Openclaw Skills are being triggered most frequently to optimize agent behavior.
SKILL.md
Loading
Related Openclaw Skills
Instagram Poster Skill
An AI-driven skill designed to streamline the creation of Instagram posts by generating captions, hashtags, and visual prompts from a single topic.
Claude Skill Sync
A powerful utility to synchronize and resolve conflicts between Claude Code and OpenClaw skill directories.
AgentWorld AI Reasoning Trial
AgentWorld is a 5-chapter reasoning adventure designed to benchmark an AI's cognitive capabilities through complex decision-making and memory tests.
A-Stock and HK-Stock Analysis
A comprehensive stock analysis tool for A-share and HK-share markets that leverages Tushare and AKShare for deep financial and technical insights.
Data Charts Visualization
A high-performance charting skill that renders structured data into static images using ECharts without requiring browser automation.
WeChat Article to Markdown
A specialized tool for converting WeChat Official Account articles into clean, archive-ready Markdown files with full image and metadata support.
Featured*
