rnd-buildlisted

# R&D Framework: Build Determine the RND artifacts directory: ```bash RND_DIR=$("${CLAUDE_PLUGIN_ROOT}/lib/rnd-dir.sh") ``` Read the plan from `$RND_DIR/protocol.md`. Check `TaskList` to identify current task states. When displaying blocked tasks or blocked-by references, always translate Claude Code internal IDs (`#<n>`) to pipeline IDs (`T<n>`) by matching against `metadata.pipelineId` or extracting the `T<n>` prefix from the blocking task's subject. If $ARGUMENTS is empty (user ran `/rnd-framework:rnd-build` with no arguments): - Read `$RND_DIR/protocol.md` and inspect `TaskList` to find the next wave where all tasks are `pending` and unblocked. - If a clear next wave is found, use `AskUserQuestion` to confirm: "Build wave-N next? (N tasks: T1, T2, ...)" with options "Yes, build wave-N (Recommended)" and "Choose a different task or wave". - If no pending wave is found, report the current state and use `AskUserQuestion`. - If the plan does not exist, prompt the user to run `/rnd-framework:rnd-start` first. If $ARGUMENTS specifies a task ID (e.g., "T3"): - Build that one task. If $ARGUMENTS specifies a wave (e.g., "wave-2"): - Build ALL tasks in that wave sequentially. If $ARGUMENTS is "next": - Use `TaskList` to find the next unblocked wave and build it. ## Build Process **For each task, spawn a Builder agent:** ``` Agent({ description: "Build task T<id>", subagent_type: "rnd-framework:rnd-builder", mode: "acceptEdits", prompt: "Task: T<id>\nRND_DIR: <path