xiaolai Anthropic Pulse — Router
| Field | Value |
|---|---|
| News index | anthropic.com/news |
| Research index | anthropic.com/research |
| Refresh cadence | daily (matrix workflow) |
| Content model | digest only — title + date + URL + 1-2 sentence summary. Click through for full content. |
This skill is auto-updated daily. The pipeline scrapes the two index pages, extracts the most recent ~20 items each, and rewrites the surface files below. Unlike the other 7 skills, this one has no docs-snapshot (the upstream is HTML, not markdown), no LLM agents (deterministic render), and no token cost per run.
Dispatch table
| Surface file | Read when the user asks about… |
|---|---|
SKILL-news.md |
Recent product launches, partnerships, region openings, model releases, policy updates, business announcements, "did X just launch?", "what did Anthropic announce recently?" |
SKILL-research.md |
Recent research papers, alignment research, evaluations, benchmarks, "what's Anthropic's research on X?", interpretability work, the Anthropic Institute / Anthropic Economic Index |
When unsure which file to read
- Word "research" / "paper" / "benchmark" / "evaluation" / "alignment" / "interpretability" →
SKILL-research.md - Words "announce" / "launch" / "release" / "partnership" / "available" / a model name →
SKILL-news.md - Generic "what's new" → read both
What this skill ISN'T
- Not a deep reference. Each digest entry is just title + date + URL + summary. For the body of a news post or a paper, WebFetch the linked URL — that's the design.
- Not historical archive. Only the latest ~20 items per feed. Older items live at anthropic.com; the freshness budget is "what changed recently."
- Not Anthropic-internal docs. This is the public-facing news + research feed. For product docs use the other 7 skills.
Auto-updated daily from anthropic.com/news and anthropic.com/research. File issues at xiaolai/anthropic-docs.