| name | find-candidates |
| description | Find, look up, research, and summarize candidates in Superposition. Use for "find Python engineers from Google", "who are the top people we haven't contacted yet", "give me a quick summary of this person before I decide", "research this candidate and check their GitHub", "what's his comp expectation / notice period / timezone", "what's her LinkedIn", or "who failed the technical screen". Also use to assess a candidate against the bar — "why does he clear the bar?", "is she really top-tier? what are the proof points?", "show me the best matches and filter out weak schools/companies". This is the read/research/assess entry point — for taking an action on a candidate use pipeline-actions; for recording a new fact use record-candidate-insights; for telling the client use pitch-to-client. |
Find & research candidates
The #1 thing operators do is look people up — to search, to get a quick read before deciding, or to
answer a factual question. This skill is read-first; only reach for invoke_agent when the user
wants active research/enrichment beyond what's already on file.
Steps
- Search or resolve.
- Open-ended ("Python engineers from Google", "ML folks") →
search_candidates with filters
(skills, company, jobTitle, school, summary, stages). Defaults to the active pipeline;
pass includeEligible/includeArchived when the ask implies it.
- A named person or LinkedIn URL →
search_candidates (pass the URL as searchTerm) then
get_candidates for full detail. Batch multiple IDs in one get_candidates call.
- Answer factual questions from what's on file first.
get_candidates (FitCheck, traits,
location, links), search_insights (person facts: visa, comp, availability), and
get_candidate_email_thread (what they last said) usually have the answer — don't guess, don't
escalate.
- For "research / summarize / dig deeper before I decide": if the on-file data isn't enough,
invoke_agent to actively research and summarize:
"Research candidate <id> (check their public profiles / GitHub as relevant) and give a concise
fit summary for job <id>."
- Present concisely with the clickable link
https://app.superposition.ai/c/{candidateId}:
highlights, fit signal, watch-outs. For lists, a compact ranked table.
Guardrails
- Respect candidate privacy — share what the task needs, not everything.
- Distinguish "find/research" (this skill) from "act on" (pipeline-actions) and "record a fact"
(record-candidate-insights); if the user shifts intent, hand off to the right skill.
- Don't invent facts; if something isn't known, say so (and offer to research).
Output
Top uncontacted for Senior Backend (5):
| Candidate | Why | Signal |
|---|---|---|
| [A. Rivera](https://app.superposition.ai/c/abc) | ex-Stripe, distributed systems | strong |
| [J. Okoro](https://app.superposition.ai/c/def) | 6y Go, fintech | strong |