What problem does it solve?
Picoclaw AI gateway traffic currently has no standardized way to detect outbound secret exfiltration or inbound command-injection payloads, leaving operators blind to data leaks and tool-abuse attempts in lightweight agent runtimes.
Core Features & Use Cases
- Traffic Monitoring Specification: Defines a detector core, Picoclaw adapter, and operator interface for HTTP/HTTPS proxy inspection with a normalized JSONL finding schema.
- Exfiltration and Injection Detection: Specifies patterns for AI API keys, AWS credentials, private keys, pipe-to-shell commands, and reverse shell shapes, with mandatory snippet redaction.
- Posture Integration: Exports a signed profile fragment so picoclaw-security-guardian can drift-check monitor status, config hashes, and log digests.
- Use Case: A builder implementing Picoclaw runtime security uses this spec to ship an opt-in, detect-and-log proxy monitor that redacts secrets before writing local findings.
Quick Start
Ask your agent to install the picoclaw-traffic-guardian skill from prompt-security/clawsec and review SPEC.md before implementing the monitor.