Publish comparison page: agentfactory vs Agent Teams vs Gas Town vs claude-squad

[stempeck]

Stage 1. The README's current comparison table (LangGraph / CrewAI / Claude Code subagents) skips every competitor the category actually discusses.

Evidence

• The above/beside-Agent-Teams conversation is being written without us: "Move over Gas Town, Claude Has First-Party Agent Orchestration" [verified: https://www.alilleybrinker.com/blog/move-over-gas-town/] and the Dynamic Workflows HN thread's category roundup names four tools, none ours [verified: https://hn.algolia.com/api/v1/items/48313051]
• Verifiable Agent Teams limitations for an honest table: experimental flag, single-machine scope, no teammate session resumption, no worktree isolation, ~7x token cost in plan mode. [verified: https://code.claude.com/docs/en/agent-teams.md, https://code.claude.com/docs/en/costs.md]

Acceptance criteria

• docs/comparison.md with rows: agentfactory, Agent Teams, Gas Town, claude-squad, multiclaude; columns: workflow definition, machine scope, crash recovery, isolation, gating, cost visibility, maturity/stars
• Honest-losses column included (Agent Teams: first-party, zero-install, free of our 7-prereq setup; Gas Town: 17.5k stars of community; we are 3 stars and AGPL) — the honesty is what makes it citable
• README table gains an Agent Teams row linking to the full page
• Every cell sourced to a URL in the page footer

Filed by the product-manager cycle af-147a1678; approved by operator: #98 (comment)