| name | preflight |
| description | Pre-flight validation checklist before executing pipeline tasks |
| user-invocable | true |
| disable-model-invocation | false |
Pre-Flight Checklist
Run before executing any pipeline script, API extractor, or statistical analysis.
Context Scan (30 seconds)
- Read
domain-facts.md — check for relevant patterns, traps, API notes
- Scan MEMORY.md "Current State" — any blockers for this task?
- If touching pipeline code: check KNOWN_ISSUES.md for the affected module
- Grep codebase for existing import/usage patterns of any libraries involved
Verify Assumptions (1 minute)
- Read the script to run — verify imports, input files, output paths
- Verify input parquets exist:
ls -la 01_Data/processed/<file>.parquet
- If API calls: verify key is set
- If numerical data: spot-check for inf/NaN in input data
Present Plan
- Write numbered plan with steps, files, commands, duration estimate, risks
- STOP and wait for user approval
Smoke Test
- Run with --sample 1 (NOT --sample 50) or verify 1 cached file
- Confirm output schema before scaling up