Idea Session Management
Use /pi to brainstorm ideas, save them as idea.md files, and pick them up days or weeks later. No idea gets lost.
The Problem
You have three ideas brewing:
- A CLI dashboard for monitoring workers
- A Slack integration for task notifications
- A caching layer for the knowledge search
Each needs research and discussion before committing to implementation. You can't do them all at once.
Start an Idea Session
/pi "CLI dashboard for monitoring CQ workers in real-time"CQ enters brainstorming mode:
- Searches existing knowledge for related work
- Runs web searches for prior art (blessed, k9s, lazydocker)
- Asks probing questions: "Who's the primary user? What metrics matter?"
After 5-10 minutes of discussion, the idea crystallizes:
✓ Saved: .c4/ideas/worker-dashboard.md
- Problem: No visibility into worker status without checking logs
- Target: Solo developers running 3+ workers
- Core: Real-time task progress, worker health, queue depth
- Risk: TUI library choice (bubbletea vs tview)Come Back Later
Next week, you want to pick up the Slack integration idea:
/pi "Slack notifications for CQ tasks"CQ finds your existing ideas:
Found 1 existing idea:
- worker-dashboard.md (7 days ago) — CLI dashboard for worker monitoring
This is a new idea. Starting fresh brainstorm...After discussion:
✓ Saved: .c4/ideas/slack-notifications.mdList All Ideas
ls .c4/ideas/worker-dashboard.md (2026-03-22)
slack-notifications.md (2026-03-29)Each file contains the full context: problem, target user, scope, risks, and research findings.
Resume an Idea
Pick up where you left off:
/pi "let's continue the worker dashboard idea"CQ loads worker-dashboard.md and resumes the conversation with full context — no re-explaining.
Promote to Implementation
When an idea is ready:
Choose next step:
[1] Auto-implement — plan → run → finish (fully automatic)
[2] Plan only — create tasks, you decide when to runOption 1 triggers the full pipeline:
/pi → idea.md → /plan → /run → /finishNo manual handoff. The idea flows straight into code.
Idea File Format
Each idea.md captures:
# Worker Dashboard
## Problem
No real-time visibility into CQ worker status...
## Target User
Solo developers running 3+ parallel workers...
## Core Concept
TUI dashboard showing task progress, worker health, queue depth...
## Research
- bubbletea: Go TUI framework, used by charm.sh tools
- k9s: Kubernetes TUI — good UX reference
...
## Risks
- TUI library lock-in
- Terminal compatibility across OS
...
## Requirements (EARS)
- WHEN user runs `cq dashboard` THEN shows live worker grid
- WHEN a worker fails THEN row turns red within 2 seconds
...Next Steps
- Feature Planning — the full /pi → /plan → /run workflow
- Knowledge Loop — how CQ learns your preferences over time