Regex-based PII redaction for LLM prompts and tool outputs, with reversible placeholders.
A fast, secure tool to redact sensitive information from shell command history
PII and secret redaction for rlg records. Built-in regex patterns for credit cards, JWTs, OAuth bearers, emails, and IPv4 addresses, plus a custom-pattern API. Sits between `Log::fire()` and the sink so logs emitted to disk / OTLP / syslog are scrubbed.
Wire-level prompt introspection for LLM SDK calls. See exactly what was sent, with credentials redacted by default.
Agent runtime: event bus, sessions, plugin trait, heartbeat, A2A delegation.
Fast, zero-dependency redaction of secrets and PII (emails, credit cards, IPs, JWTs, API keys, and more) from text and logs.
A simple secret scanner and redactor utility
Safety layer for Claude Code: classifier-gated bash hook + MCP server blocking dangerous command compositions.
Behavior contracts for AI agents — tested in your PR, enforced at runtime. Core engine: parser, writer, content-addressed store, replay, and nine-axis behavioral differ.