Problem
After thinktank run starts agents, the terminal goes silent for 5–20 minutes. The only feedback is a single line: "Running N agents in parallel (claude-code)...". With --verbose the raw agent output floods the terminal. There is no middle ground.
Users have no way to know:
- Which agents are still running vs. finished
- Whether an agent is stuck or making progress
- How long until the run completes
Proposed Solution
Add a live status panel that updates in-place (using ANSI cursor movement) showing per-agent state:
Running 5 agents in parallel (claude-code, sonnet)...
Agent 1 ██████████████████░░ running 1m 23s
Agent 2 ████████████████████ done 1m 10s ✓ tests passed
Agent 3 ████████████████████ done 1m 45s ✗ tests failed
Agent 4 ██████░░░░░░░░░░░░░░ running 42s
Agent 5 ████░░░░░░░░░░░░░░░░ running 28s
States: waiting, running, done, timeout, error
When NO_COLOR or non-TTY, fall back to a simple log-style output:
[1m 23s] Agent 1: still running...
[1m 10s] Agent 2: complete (tests passed)
Acceptance Criteria
Problem
After
thinktank runstarts agents, the terminal goes silent for 5–20 minutes. The only feedback is a single line: "Running N agents in parallel (claude-code)...". With--verbosethe raw agent output floods the terminal. There is no middle ground.Users have no way to know:
Proposed Solution
Add a live status panel that updates in-place (using ANSI cursor movement) showing per-agent state:
States:
waiting,running,done,timeout,errorWhen
NO_COLORor non-TTY, fall back to a simple log-style output:Acceptance Criteria
NO_COLORis set or stdout is not a TTY (e.g., CI)--verbosestill dumps raw agent output (current behavior preserved for debugging)