agent-office v0.1.1 — Demo, CI, and safer contribution

This release makes the two-verb idea easier to see, try, and trust.

Highlights:

• Adds a lightweight README animation and the full 720p demo with audio.
• Adds bilingual 30-second setup, tmux lifecycle, reset, and troubleshooting guidance.
• Adds isolated fake-tmux smoke tests on macOS and Ubuntu plus ShellCheck.
• Adds a security policy, private vulnerability reporting, issue routing, and a pull request checklist.

The runtime contract is unchanged: agent-office still has exactly two verbs,
up and send. It still uses the isolated agent-office tmux server and never
reads other terminals.

Verified commit: 17165a7

CI: ShellCheck, macOS, and Ubuntu passed

agent-office v0.1.0 — Two verbs, real tmux rooms

agent-office v0.1.0

The first public release of a tiny, terminal-native delegation layer for local CLI agents.

What ships

• up starts real CLI agents in named rooms on the isolated tmux -L agent-office server.
• send types one task into one room or broadcasts it to all agent rooms.
• Bare up auto-detects claude, codex, and grok, then adds a spare sh room.
• name:command rooms support other interactive terminal CLIs.
• send --all skips the spare shell unless --include-shell is explicit.
• English and Simplified Chinese documentation, a reusable agent skill, issue templates, and an MIT license.

Deliberate boundaries

• No API proxy.
• No model adapters.
• No workflow engine.
• No terminal scraping or hidden result collection.

Requirements

• Bash
• tmux
• At least one local CLI agent already installed and authenticated

Quick start

chmod +x agent-office.sh
./agent-office.sh up
./agent-office.sh send --all 'State your role and wait for a focused task.'

The agent in your current conversation remains the lead. The other agents stay in their own real terminal rooms.