Agentctl v0.5.0

@lab42/agentctl v0.5.0

What's new since v0.4.0:

Features

• JSON Schemas for .ai/ files — config.yaml, permissions.yaml, and mcp.yaml now have JSON Schemas, so editors (VS Code, JetBrains) get autocomplete, inline validation, and hover docs. Stub files ship with a # yaml-language-server: $schema=... comment as a fallback, and the schemas are pending registration in SchemaStore for zero-config detection. (#37, #53, #54)
• Cross-runtime consistency audit — agentctl audit checks that a command resolves to the same permission decision across every enabled runtime and flags divergences by severity (critical/warning/info). (#47)
• Explain permission decisions — agentctl explain <command> shows how a shell command is evaluated under deny_over_allow, including which pattern matched and why. (#46)
• Pre-commit hook setup — agentctl init --with-hooks auto-detects your hook manager (lefthook, husky, simple-git-hooks, or a raw .git/hooks fallback) and wires up agentctl check so drift is caught before commit. (#52)
• Better error messages — parse and validation failures now cite the file, line, and a contextual snippet with a fix suggestion. (#51)

New commands

• agentctl audit — report cross-runtime permission divergences (exit 1 on critical)
• agentctl explain <command> — show the permission decision and matched pattern for a command
• agentctl init --with-hooks — set up a pre-commit hook running agentctl check

Docs

• init now promotes npx @lab42/agentctl as the primary install path. (#50)
• Added a Code of Conduct.

Upgrading

No breaking changes. Generated output is unchanged from v0.4.x.

pnpm add -D @lab42/agentctl@0.5.0