Claude Code guardrails v1.0.2

• The two skills pre-approve only guardrails' own command (node ${CLAUDE_PLUGIN_ROOT}/bin/claude-guardrails.mjs ...) instead of any node command, so while one of them runs, other Node code still asks first.
• An icon for the plugin's listing in Anthropic's plugin directory, and the listing's links in plugin.json: documentation, support, privacy and terms.
• README: a Privacy section (what the hook reads, what it writes, and that it sends nothing anywhere).

See the README to set it up, and CHANGELOG.md for every version.

Claude Code guardrails v1.0.1

• --dry-run now passes for poetry, uv, flit, hatch, vsce and ovsx publishes too, as the README said.
• The reasons for docker-prune and aws-destroy name what they catch.
• README: what the rules and your own patterns really cover (.npmrc at commit, the Docker and AWS lists, deny and ask patterns on the whole command line), measured reason lengths and timings, and Node 18 in CI.

See the README to set it up, and CHANGELOG.md for every version.

Claude Code guardrails v1.0.0

• First release: a tokenizer for bash, PowerShell and cmd, rules for destructive deletes, force pushes, lost work, secrets in commits, SQL drops, publishing, infrastructure destroys, remote scripts, secrets files, lockfiles and writes outside the project; presets strict, balanced and relaxed; self-protection; project and user installs; a plugin.

See CHANGELOG.md for every version, and the README to install.