v0.1.0 — first formal release

First formal release of agentfactory — a multi-agent orchestration CLI for Claude Code — consolidating ten weeks of development (snapshot tags V001–V012).

What it does

Turn SKILL.md files into autonomous agents that execute structured workflows:

• Formulas — declarative TOML workflows (steps, DAG dependencies, variables, gates) kept separate from agent personas; /formula-create generates one from a SKILL.md, af formula agent-gen turns it into an agent
• Crash & context-compression recovery — af prime re-injects identity and current step on every prompt; checkpoints and PID-stale-lock detection let a crashed agent resume where it stopped
• Inter-agent mail — issue-store-backed messaging with broadcast groups, reply threading, and hook-driven delivery
• Autonomous dispatch — GitHub issue/PR label matching with issue→PR handoff and cycle locking
• tmux-native sessions — every agent is an inspectable, attachable Claude Code session (af attach <agent>)
• 19 shipped formulas — implementation, design, review, root-cause, and utility agent families
• Optional web console — loopback-only control room: floor view, task slinging, dispatch status, browser formula authoring

Requirements

Go 1.24+, Python 3.12, tmux 3.0+, jq, Claude Code CLI. Install from source: make build && make install. Prebuilt binaries are on the roadmap (#84).

See CHANGELOG.md for the themed change history and docs/ for the formulas reference, agent lifecycle, and recovery model.