SentiClaw

Runtime AI security for OpenClaw agents. Protects against prompt injection, identity spoofing, PII leakage, and runtime abuse. Drop-in 6-layer security middl...

SentiClaw appears to do what it says, but check a few operational details before installing: 1) Ensure the host has the 'openclaw' CLI (or update AuditLogger) and that you trust the CLI's configured credentials — alerts use that CLI to send messages. 2) Choose and secure the audit_db_path (default senticlaw_audit.db) and set file permissions to limit access to the stored events. 3) Set alert_channel and alert_channel_id deliberately (avoid exposing alerts to untrusted endpoints). 4) Review the code path that calls subprocess.run (audit._alert) if you need stricter controls on outbound telemetry. 5) Run the included tests locally (tests/run_tests.py) and inspect logs before deploying in production. If you require stricter guarantees, consider running the skill in a sandboxed environment with restricted network access and review any OpenClaw credentials that the host CLI uses.

Type: OpenClaw Skill Name: senticlaw Version: 1.0.0 SentiClaw is a comprehensive security middleware for OpenClaw agents designed to mitigate prompt injection, PII leakage, and unauthorized access. The bundle implements a 6-layer defense strategy including identity verification, input sanitization, outbound content gating, PII redaction, rate limiting, and path/URL validation. It features a legitimate auditing system using SQLite and an alerting mechanism that utilizes the 'openclaw' CLI to notify users of detected threats. No evidence of malicious intent, data exfiltration, or unauthorized persistence was found; the code logic is transparent and strictly aligned with its stated purpose of enhancing agent security.