---

name: pipeline-authoring description: "Author, validate, debug, and adapt RocketRide .pipe workflows." metadata: disco-role: operating disable-model-invocation: true license: MIT

Pipeline Authoring

Use this sub-skill when you need to create or revise a RocketRide .pipe workflow, reason about data-lane versus control/invoke wiring, or prepare a workflow for static validation before an engine run.

Use this for

• New .pipe workflows or edits to existing ones
• Repairing lane wiring, source/target placement, or control connections
• Replacing hardcoded credentials with environment placeholders
• Adapting RAG, document-processing, multi-agent, or n8n round-trip patterns
• Checking a workflow before handing it to the engine or IDE

Do not use for

• SDK method details, client usage, or CLI commands → ../sdk-clients/SKILL.md
• Provider catalog, node READMEs, or schema depth → ../nodes-catalog/SKILL.md
• Server startup, deployment, or runtime operations → ../runtime-deployment/SKILL.md
• Deeper MCP or n8n integration behavior → ../mcp-and-integrations/SKILL.md

Working order

1. Read the current pipeline shape and identify the source, targets, and any nested sub-pipelines.
2. Separate data lanes from control/invoke connections.
3. Replace secrets with ${ROCKETRIDE_*} placeholders.
4. Keep project_id, source, viewport, and version aligned with the file's actual entry point and canvas state.
5. Run static validation before any engine run.

Validation helper

Use the shared static probe ../../scripts/rocketride_static_probe.py to check JSON shape, references, lane wiring, and control ownership before execution.

Reference map

• Schema and recipes
• Troubleshooting

What good output looks like

• Unique component ids
• One explicit entry source
• Correct lane types end to end
• Control connections on the controlled node
• No hardcoded secrets
• Safe, engine-ready JSON