Layoff Signal

HOW in lib/layoffsignal.py. Index: reference/docs/layoffsignal/raw/endpoints.md.

Kind: hyreflow Native — read this first

• First-party capability, no external vendor. Built on free public RSS (Google News + per-site feeds).
• No BYOK, no env key. A native has no third party to bring a key for → it runs on hyreflow infra and is metered in hyreflow credits — 0 credits under current pricing. The provider-precedence waterfall does not apply.
• ToS-clean. Surfaces facts from primary public reporting (the same sources a human tracker reads). Does not scrape or resell anyone's curated database (e.g. layoffs.fyi's Airtable). Attribute the originating publisher. Use as a signal/trigger, not a redistributed dataset for customers to browse.

Auth & config

• Base: https://news.google.com/rss · No authentication (native; no env var).
• Instantiate LayoffSignal(). Discovery returns links only — fetch article bodies with the firecrawl adapter (it also resolves the Google News redirector).

Operations

• Broad sweep: layoff_news(when="7d") → Google News RSS with LAYOFF_TERMS OR-joined. Verified live (2026-06-01) returning 50 dated articles (Meta, etc.).
• Targeted employer: layoff_news(company="Microsoft", when="30d") — the poach trigger.
• Custom query: google_news('"layoffs" site:techcrunch.com', when="7d") — full Google operators.
• Clean direct links: site_feed(url) / site_feeds() over curated outlets (DEFAULT_SITES) — no redirector to resolve. Bloomberg/Reuters are paywalled → rely on google_news to surface those.
• Whole-stream: iter_layoff_news(when="7d", companies=[...]) — broad + per-company + site feeds.
• Resolve: resolve_link(url) best-effort follows a redirector → publisher URL (prefer firecrawl).

Query operators (Google News RSS)

q accepts "exact phrase", OR, site:, intitle:, -exclude, plus News-native recency via the when= arg (7d/24h/1h). layoff_news builds "layoffs" OR "job cuts" OR "cutting jobs" OR … when:7d.

⚠️ Two gotchas

1. Redirector links. google_news/layoff_news links are news.google.com/rss/articles/CBMi… (encoded redirect), not the publisher URL. Resolve via resolve_link() (best-effort) or — more reliably — hand the link to firecrawl, which follows it and returns the article body.
2. Prefer per-site feeds for clean links. site_feed() gives direct publisher URLs.

Why it's here — the signal model (the product angle)

Layoffs are a live recruiting trigger: a company cutting staff = a pool of available candidates to poach

• a BD opening with the survivors. As a native it runs on free public RSS and costs 0 credits under current pricing, so sweeping broadly is safe — spend starts downstream, on enriching the people you keep. See reference/provider-precedence.md (natives are exempt) and the recipe below.

Handoff (the recipe)

recipes/layoff-signal-to-poach.md: sweep feeds → resolve+fetch article bodies (firecrawl) → Claude extracts {company, headcount, %, date, location, source} (free, Model A) → dedupe → resolve company → source affected roles (github for eng, apollo/aiark otherwise) → enrich (waterfall) → outreach/ATS.

Callable surface — lib/layoffsignal.py

Import: from lib.layoffsignal import LayoffSignal → instantiate LayoffSignal() (reads key from env). Base: https://news.google.com/rss. Generic passthrough: request(method, path, *, params, json).

• google_news(query: str, *, when: str | None = None, hl: str = 'en-US', gl: str = 'US', ceid: str = 'US:en', limit: int | None = None) -> list[dict] — GET /search?q={query} — Google News RSS search (free, no key). when adds a recency
• iter_layoff_news(*, when: str = '7d', companies: list[str] | None = None, include_sites: bool = True, **kw) -> Iterator[dict] — Walk a full sweep as one stream: the broad layoff query + one query per company (if given)
• layoff_news(*, when: str = '7d', company: str | None = None, terms: list[str] | None = None, **kw) -> list[dict] — Convenience: google_news() pre-built with LAYOFF_TERMS (OR-joined). Pass company to scope
• resolve_link(url: str) -> str — Best-effort: follow a Google News redirector link → the real publisher URL (returns the
• site_feed(feed_url: str, *, limit: int | None = None) -> list[dict] — GET {feed_url} — fetch ANY publisher RSS/Atom feed → items with CLEAN direct links (no
• site_feeds(feeds: dict | None = None, *, limit_each: int | None = None) -> dict[str, list[dict]] — Fetch several publisher feeds → {feed_name: [items]}. Defaults to DEFAULT_SITES. Per-feed