What problem does it solve? It automates importing and smoke-testing a Twitter/X network into the Powerpacks pipeline, resolving handles through RapidAPI crawling, OpenAI-based MOE enrichment, and LinkedIn validation without manual orchestration. ## Core Features & Use Cases - Manifest-only idempotent runs: Each handle writes a fixed output directory with a single manifest.json, and reruns skip steps whose output already exists so spend is never duplicated. - Spend-gated execution: RapidAPI crawling and OpenAI/LinkedIn resolution run only with --approve-spend; without it the run stops and emits a needs_approval manifest naming the step and estimated calls. - Use Case: Run a 10-row sanity import for a handle like example_operator to validate the pipeline end to end, then approve spend to produce people.csv and rebuild the merged local network index. ## Quick Start Use the import-twitter skill to run a 10-row smoke import of the Twitter handle example_operator and summarize how many profiles resolved to LinkedIn URLs.