conference-scoutlisted
Install: claude install-skill scicrow/conference_connector
# conference-scout
You are driving `conference_connector` (a separate installed library: `recon` / `ingest` /
`prefilter` / `validate` / `rank` / `render`) through a project directory that holds
one user's configs, cached data, and outputs for one conference. Everything
downstream of the item list is arithmetic the library already does correctly -- don't
re-derive it. Everything upstream of that is judgement: this file is about the parts
where you have to look at real pages, real text, and real names and decide something.
## 0. Before anything else: is there a project directory?
Look for `config/config.yaml` in the current directory. If missing, this is a new
project. **Do not hand the user a blank YAML file and ask them to fill it in --
interview them and write it yourself.** That's the whole point of driving this
through a skill rather than a README: the user describes their work and goals in
conversation, you translate that into the structured file the library needs.
1. Interview the user. You need: what conference, what URL, what they're trying to
get out of it (find a collaborator? a host lab? scout competitors? just triage a
big programme?), and enough about their own work to write real keyword lists --
not just a topic name, but the vocabulary someone else might use for the same idea
with different words. Weak, generic keywords are the single most common way a good
candidate silently never gets read.
2. Run `conference_connector init` to scaffold the