| name | import-twitter |
| description | Import or smoke-test Twitter/X network artifacts through the RapidAPI + MOE + LinkedIn resolution pipeline. Use for $import-twitter. |
import-twitter
Use this skill for $import-twitter or Twitter/X network import testing.
This is an alias/wrapper around discover/twitter/network_import.py. It is
manifest-only: one idempotent run per handle writes the fixed dir
.powerpacks/network-import/discover/twitter/<handle>/ (overwritten in place)
plus a single manifest.json. The RapidAPI-backed crawl and the OpenAI/RapidAPI
resolution steps are spend-bearing and run only with --approve-spend.
10-row sanity command
Without --approve-spend, run stops at the first spend step and emits a
needs_approval manifest naming the step + estimated calls:
uv run --project . python packs/ingestion/primitives/discover/twitter/network_import.py run \
--handle example_operator \
--max-pages 1 \
--limit 10 \
--min-score 0
Approve the spend by re-running the same command with --approve-spend; it
advances the whole pipeline in one pass and never re-spends steps whose output
is already on disk:
uv run --project . python packs/ingestion/primitives/discover/twitter/network_import.py run \
--handle example_operator \
--max-pages 1 \
--limit 10 \
--min-score 0 \
--approve-spend \
--linkedin-workers 10 \
--aggregator-workers 10
Inspect progress any time from the manifest (no spend):
uv run --project . python packs/ingestion/primitives/discover/twitter/network_import.py status --handle example_operator
Notes
- Requires a RapidAPI Twitter key subscribed to
twitter241 (RAPIDAPI_TWITTER_KEY or RAPIDAPI_KEY).
- MOE uses OpenAI and is spend-gated behind
--approve-spend; do not run deep research for this smoke.
- RapidAPI LinkedIn validation uses
RAPIDAPI_LINKEDIN_KEY or RAPIDAPI_KEY and is spend-gated behind --approve-spend.
- Resume is by artifact: a rerun skips steps whose fixed output CSV is already present, so it never re-crawls or re-spends. Delete the handle dir to force a clean re-run.
- Final output should include
people.csv; summarize as x/10 linkedins plus counts, not raw rows.
After Twitter finishes, rebuild the merged network and local index with the
indexing fan-in:
uv run --project . python packs/indexing/primitives/index_contacts_pipeline/index_contacts_pipeline.py run --operator-id <operator-id>.
