planning-in-backgroundlisted

# Background Planner Multi-LLM parallel planning with context-safe auto-save. ## Quick Start ```bash # 1. Parse topic and perspectives # 2. Create: .context/plans/ # 3. Determine round: R01, R02, ... # 4. Run agents in background → {round}-{agent}.md # 5. Guide user to merge when ready ``` ## Output Convention ``` .context/plans/ ├── R01-claude.md # Round 1: Claude's plan ├── R01-codex.md # Round 1: Codex's plan ├── R01-gemini.md # Round 1: Gemini's plan ├── R01-merged.md # Round 1: merged plan ├── R02-claude.md # Round 2: refined after feedback └── R02-merged.md ``` **Round number** increments each planning iteration: ```bash mkdir -p .context/plans ROUND=$(printf "R%02d" $(( $(ls .context/plans/R*-*.md 2>/dev/null | sed 's/.*\/R\([0-9]*\)-.*/\1/' | sort -rn | head -1 | sed 's/^0*//') + 1 ))) ``` ## Provider Selection | Provider | Best For | Command | |----------|----------|---------| | **Claude** | Complex analysis, architecture, deep codebase reading | `Task({ run_in_background: true })` | | **Codex** | Technical specs, code design | `nohup codex exec --full-auto -o {output.md} "prompt" > log 2>&1 &` | | **Gemini** | Creative ideas, UX design, long docs | `nohup gemini -p "prompt" -o text > {output.md} 2>/dev/null &` | | **Ollama** | Sensitive data, local | `ollama run llama3.2` | > **Gemini v0.26+**: Use `-p "prompt"` for non-interactive, `-o text` for clean output, `--yolo` for auto-approve file writes. Redirect stdo