MetalHexx/rad-execute

Step 1: Initialize

You are an orchestrator. You'll be using the rad-orchestration skill for this project. Read the skill and prepare to use it to run the execution pipeline.

Step 2: Approve plan

The Master Plan is complete. As a human reviewer, I have approved the plan and am ready to execute. Mark the plan as approved and begin execution of the project.

Step 3: Source Control Initialization

Before the first pipeline tick, ensure pipeline.source_control is populated in state.json. The commit and PR gates read from this state — without it, the walker halts when it reaches either conditional.

1. Run node {skillRoot}/scripts/gather-context.js --project-name {PROJECT_NAME} and parse its JSON output. {skillRoot} is the directory containing this SKILL.md; {PROJECT_NAME} comes from the /rad-execute argument or conversation context.
2. If sourceControlInitialized === true, skip to step 4 — resume is ceremony-free.
3. Otherwise, follow references/source-control-init.md to resolve each init field (prompting only when needed) and fire the source_control_init pipeline event.
4. Proceed with execution.

Step 4: Execute Plan

Execute the project according to the approved Master Plan using the proper execution pipeline.

Resuming a project: to resume execution (or determine the next pending action on a fresh session), fire the pipeline start event:

node {orchRoot}/skills/rad-orchestration/scripts/pipeline.js --event start --project-dir {projectDir}

start returns the current pending action without mutating state, so it is the correct entry point both for first-time execution and for resume. Do not invent event names like tick, next_action, or get_next_action — they will fail with Unknown event.

Step 5: Pipeline Error Handling

• If any errors occur with the pipeline during execution, use the rad-log-error skill to log them
• Do not try to fix the pipeline code, simply work around it.
• Ensure that error messages are clear, actionable, and include relevant information about the failure point.