leak-check
by khaney64
Coding Agents & IDEsBrief
Scan session logs for leaked credentials.
Description
What This Skill Does
Scans OpenClaw session JSONL files for credentials that may have been sent to AI providers. Matches credential fragments against configurable patterns and identifies which provider (Anthropic, OpenAI, Google, etc.) received the data. Distinguishes real leaks from config echoes where the config file itself was read during a session.
Runs entirely local with no network access, no dependencies, and no file writes, so it can be trusted to scan sensitive session data without risk of further exposure.
When to Use It
- Checking if an API key was accidentally sent to an AI model
- Auditing past sessions after a suspected credential exposure
- Verifying no secrets leaked before sharing session logs
- Running a post-incident scan across all stored sessions
- Monitoring which AI providers received sensitive data
Example Workflow
User asks: check if any of my credentials leaked in recent sessions
- 1
Read leak-check.json config from ~/.openclaw/credentials/ to load credential fragments
- 2
Scan all JSONL files in ~/.openclaw/agents/main/sessions/
- 3
Match each session's content against configured wildcard patterns
- 4
Distinguish config echoes (where config file was read) from real leaks
- 5
Report matched credentials with session ID, timestamp, and provider name
Discord-formatted report listing 2 leaked credentials with session IDs and providers, or a clean confirmation if none found
Requirements
- leak-check.json config file with credential fragments to search for
Installation
clawhub install khaney64/leak-checkRun the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.