{"id":103326,"plugin_id":"plugin_asdk_app_6aedf6ea2fac819180910f8b98663c77","kind":"skill","collection_source":"plugin_package","comparison_source":null,"observed_at":"2026-11-18T06:19:20.707Z","digest":"d58c4159b35be447e97828170441e3cc50770ae3f52ccfbd3776f14600a19bb9","against":null,"payload":{"description":"Use to find net-new people/companies by employer, name or title, refine criteria and page results. Other filters route to searches-advanced.","included_files":[],"name":"searches","skill_md_contents":"---\nname: searches\ndescription: Use to find net-new people/companies by employer, name or title, refine criteria and page results. Other filters route to searches-advanced.\nallowed-tools: Read, Grep\n---\n\n# Clay search\n\nBefore using this skill, read the `clay` entry skill if you haven’t already this session, and follow its access guidance.\n\nSearch finds net-new people/companies; use `audiences` for workspace records.\n\n## Simple queries\n\nOutside Sculptor's Search panel, use these patterns without the full reference when they\ncapture **every** criterion. In that panel, read its required DSL reference before querying.\nPreserve lists/exclusions; add no unrequested filters. Otherwise read **`searches-advanced`\nbefore querying**: geography/metro/HQ versus offices, industry, growth, size, similarity,\nemployer history, hiring, audiences, education, counts, per-company limits or unknown\nfields. It routes to the full runtime reference. Never drop criteria to avoid it.\n\nSyntax: `select from people|companies where <predicate>`. Double-quote strings;\nparenthesize lists. `and` binds tighter than `or`; group alternatives. Text matching is\ncase-insensitive; `contains` matches tokens, not substrings. Omit the DSL `limit` unless\nthe user explicitly caps the total across all pages, so \"show me more\" can continue the\nsame search. For ordinary requested counts, use the executing tool's page-size `limit`\nwithin its allowed maximum; fetch additional pages only as needed.\nTranslate ordinary terms to English; preserve proper names/IDs.\n\n**Current employer + role:**\n\n```text\nselect from people where clay.filter_to_companies((\"stripe.com\")) and experiences.any(is_current = true and job_title is_similar_to (\"software engineer\"))\n```\n\nUse supplied identifiers or confidently known household-name domains. The company\nfunction belongs at the root. Keep title/seniority in **one current experience**;\nneither is a people-root field.\nFor an uncertain employer name, replace the identifier function with\n`company_name contains \"Name\"` inside that same experience; never guess a domain.\n\nPerson names use root-level `full_name contains \"Person Name\"`; no first/last-name fields.\n\nTitles default to `job_title is_similar_to (\"role\")`; use only user-supplied roles,\nnot invented synonyms. Exact titles use `=` or `in (...)`; literal keyword requests\nuse `contains`, keeping exclusions with `not`. Lists mean OR; required co-occurring\nwords need AND. Quote full phrases. Keep title predicates inside experience.\n\nOnly filter seniority when requested: senior individual contributors use\n`seniority = \"Senior\"`, not leadership. Complete experience values: `Founder`, `Owner`,\n`Board Member`, `Partner`, `C-suite`, `VP`, `Director`, `Head`, `Manager`, `Senior`,\n`Mid-level`, `Entry`, `Intern / In Training`, `Unknown`. Use `=`, `!=`, `in`, `not_in`;\nleadership tiers require leadership intent. Not job-posting values.\n\n**Specific companies:** one inclusion call, all supplied identifiers (at most 1,000):\n\n```text\nselect from companies where clay.include_company_identifiers((\"stripe.com\", \"https://www.linkedin.com/company/notionhq/\"))\n```\n\nDo not substitute `domain =`, guess domains or fabricate LinkedIn slugs. Ambiguous\ncompany names and \"companies like X\" require advanced identification guidance.\n\n## Execute and page\n\nUse the access method selected by `clay`. Discover operation schemas;\nCLI equivalents are `clay searches query-mode create --query '<query>'`, then\n`clay searches query-mode run <searchId> [--limit <n>]`. Create returns `searchId`, not\nrecords; run returns `data`, `hasMore` and optional `periodQuota`.\n\n**Forward-only iterator:** reuse its ID for the next page; no replayable cursor.\nPage size is separate from total limit.\nRead only enough pages to satisfy the current request; for an unspecified count, show\none page. On \"show more\", advance only when `hasMore` is true. Check quota before each\npage and stop near exhaustion unless the user confirms continuation. Changed criteria\nrequire a **new query retaining unchanged criteria**, not filtering the old page.\n\n## Limits and failures\n\nFollow `workflows-discover-actions/cost-and-budget.md`; keep routine checks internal.\nBefore a generic large search, suggest 2–3 supported narrowing options or a small limit\nand wait for confirmation/narrowing. Skip for bounded requests. If `periodQuota` is\npresent, ask before planned N results leave `remaining − N` below 15% of `limit`;\ncontinue only if confirmed, otherwise offer a smaller pull/stop. Skip absent quotas.\n\n### Quotas\n\nOn `quota_exceeded` (HTTP 402), stop and read `searches-advanced` → Quotas before any\nretry/report. Only an explicit per-request cap permits one smaller-page retry; never\nshrink pages or recreate searches to evade other caps. Preserve its billing/support links.\nFor validation, server or rate-limit errors, read advanced reference/retry guidance\nbefore correcting or retrying: a failed run may already have charged/counted its page.\nDo not evade authentication or approval requirements by changing transports/restarting.\n\nFor enrichment or persistence, use `routines`: inspect the paginated routine list,\nprefer managed routines and match actual input schemas. Search missing email/phone\nis not proof enrichment is unavailable. Preserve shared cost, approval and batch rules.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}