Pre-flight Validation: Don't Burn Tokens

The Problem & The Cost

• **The lesson that cost real money** — a 7-step pipeline failed at step 6 because the AEM MCP server wasn't running. Steps 1-5 consumed tokens, generated files, made progress. Step 6 hit 'connection refused.' All wasted.
• The economics — ROI of 200-500x
  • › Pre-flight check: ~$0.01 (one lightweight MCP call)
  • › Failed pipeline: ~$2-5 (multiple agents, tools, wasted output)
• **Rule of thumb** — if the pipeline has more than 3 steps, add pre-flight validation. The 2 seconds of checking save 15 minutes of failing.

What to Check & Where

1. 1 Pre-flight checklist — verify before any multi-step pipeline
   • › Do all required inputs exist? (ticket ID, spec files, URLs)
   • › Are MCP servers responsive? (make a lightweight call)
   • › Does the target branch exist? (git verify)
   • › Is the environment correct? (Node version, config files)
2. 2 We added pre-flight checks to every coordinator skill — before dispatching any subagent, the coordinator runs a quick validation. If anything fails, it stops immediately with a clear error message.
3. 3 Add a lightweight MCP call as the first step of any multi-step workflow — e.g., `getNodeContent path: /` for AEM
4. 4 Build pre-flight into coordinator skills, not individual steps — the coordinator validates once, then dispatches with confidence