Scan OpenClaw session JSONL files for leaked credential patterns and identify external AI providers.
npx skills add https://github.com/SJTU-IPADS/SkVM-data --skill leak-check
Please help me install this Agent Skill. Skill: leak-check Source: https://github.com/SJTU-IPADS/SkVM-data/tree/main/skills/leak-check Command: npx skills add https://github.com/SJTU-IPADS/SkVM-data --skill leak-check
Credentials can be leaked in OpenClaw session logs, exposing tokens and keys. This skill scans JSONL session files to identify where credentials appeared and flags which external AI provider received the data.
Run the leak-check script to scan your OpenClaw session logs for leaked credentials.
High-intent search queries and answers about installing and using this skill.
Scanning JSONL session files with the leak-check script detects leaked credentials by matching content against configured credential patterns. It distinguishes actual external provider data exposure from harmless internal delivery echoes.
Credential leak detection distinguishes real leaks from internal echoes by identifying when credential definitions appear within the JSONL session logs. It ignores these harmless internal echoes and flags actual data exposure to external AI providers.
Yes, you can track which external AI provider received leaked credentials in JSONL files. The scan resolves provider information across all stored session files, providing per-session and per-credential details including session ID and timestamp.
The scan provides per-session and per-credential details for each leaked credential found in session logs. This includes the specific session ID, timestamp, and the external AI provider information associated with the data exposure.
Yes, the results of scanning JSONL session files for credential leaks are deterministic. The leak-check script ensures consistent outcomes by applying configured patterns, session tracking, and provider resolution logic across all scans.
lemonlqfSearch and analyze JSONL conversation logs using jq and rg.
usepowershellRoute agent notifications to mapped channels using CHANNEL:<type> prefixes.
Clear-CapabilitiesRotate leaked API keys and credentials end-to-end across supported providers.
UnitOneAIAnalyze security events and generate incident reports with this tool.
