execute-planlisted

# Execute Plan > Relies on BDK foundation (STARTUP_INSTRUCTIONS.md) for project context and MCP tool preference. !`python3 ${CLAUDE_PLUGIN_ROOT}/scripts/inject.py --chain ${CLAUDE_PLUGIN_ROOT}/fragments/tool-tiers/impact.chain.json` ## Step 0: Create Task List Before executing, create TaskList from plan: - One `TaskCreate` per implementation task - Set status `in_progress` on start, `completed` when done ## Step 1: Execute the Plan $ARGUMENTS Per task: 1. **Has `**Test cases:**` section?** - YES → invoke `/bdk:test-driven-development` before any implementation. Pass test bullets + implementation desc. No implementation until GATE 2 (RED verified). - NO → implement directly (e.g. refactor-only) 2. After `completed`, check context: - **< 40%** → next task - **≥ 40%** → pause, write: > ⚠️ Context at XX% — consider running `/bdk:save-progress [slug]` before continuing. Ask: "Continue or save progress first?" Wait for response. > See `subagent-execute-plan` "Context-stop policy" for the autonomous-mode threshold (≥ 50%, higher because the coordinator's per-task state is lighter). 3. Blocked or unclear → stop, ask user ## Step 2: Verify and Fix All tasks done → inject project tools context: - Test tools: !`python3 ${CLAUDE_PLUGIN_ROOT}/scripts/get_settings.py test-tools` - Lint tools: !`python3 ${CLAUDE_PLUGIN_ROOT}/scripts/get_settings.py lint-tools` Run verification agents **in parallel**: ### 2a. Run Tests Delegate to `test-runner` agen