v1.0.0

First stable release.

Reusable guardrails for AI-assisted coding agents — egress guard, safe terminal, publish gate, workspace scanner, secret store, commit guard, memory encryption, and more.

Supports Hermes, OpenClaw, and Claude Code. Pure Bash, no external dependencies beyond jq, openssl, and git. Runs fully offline.

See the README for quick start, architecture overview, air-gapped install steps, and a case study walkthrough.

Agent Security Kit v0.1.0

Initial public release of Agent Security Kit.\n\nIncludes reusable guardrails for Hermes, OpenClaw, and Claude Code: egress guards, safe terminal hooks, publish gates, workspace scanners, secret-store loaders, memory encryption, sanitized backups, commit hygiene helpers, staging scans, self-review/context prompts, and CI smoke tests.\n\nValidation:\n- scripts/publish-check.sh passed locally\n- GitHub Actions Smoke Test passed on main