Skip to content
Skills Library
Data & analytics · Tool integration

newscatcher

/newscatcher

Use for anything about the news or current events, and for setting up or troubleshooting NewsCatcher — what happened, latest developments, coverage of a company, person, topic or event, how many articles mention something, news trends, a table of events (deals, layoffs, funding, recalls), watching companies or topics on a schedule. Maps each kind of question to the right NewsCatcher tool, explains costs, and fixes setup and error problems.

Licence
MIT
Last repository change
12.11.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
2
Scripts
1
HTML comments
0
URLs in SKILL.md
1
Deprecation

No signals detected. Repository changed on 12.11.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What to check before installing
  • 1 scripts in this package. Review them before installing, as Claude may execute them.

Scripts: scripts/render-report.mjs

What it does and when to use it

Use for anything about the news or current events, and for setting up or troubleshooting NewsCatcher — what happened, latest developments, coverage of a company, person, topic or event, how many articles mention something, news trends, a table of events (deals, layoffs, funding, recalls), watching companies or topics on a schedule. Maps each kind of question to the right NewsCatcher tool, explains costs, and fixes setup and error problems.

Original author description

Use for anything about the news or current events, and for setting up or troubleshooting NewsCatcher — what happened, latest developments, coverage of a company, person, topic or event, how many articles mention something, news trends, a table of events (deals, layoffs, funding, recalls), watching companies or topics on a schedule. Maps each kind of question to the right NewsCatcher tool, explains costs, and fixes setup and error problems.

How it works

  1. Pick the tool
  2. Costs (credits, from one balance)
  3. Before a paid run (startresearch, createmonitor)
  4. While a run is going
  5. Habits that give better answers
  6. Setup
  7. When something fails
Original SKILL.md headings · 109 lines, 1,108 words
SKILL.md preview
---
name: newscatcher
description: "Use for anything about the news or current events, and for setting up or troubleshooting NewsCatcher — what happened, latest developments, coverage of a company, person, topic or event, how many articles mention something, news trends, a table of events (deals, layoffs, funding, recalls), watching companies or topics on a schedule. Maps each kind of question to the right NewsCatcher tool, explains costs, and fixes setup and error problems."
---

# NewsCatcher

NewsCatcher covers the news in every language and runs a research agent on top of it. This skill is the map: which
tool answers which question, what it costs, and how to fix problems. Use-case skills (briefings, company research,
adverse media, trackers…) build on it.

Runs on the NewsCatcher MCP tools. If tools named `search_news`, `start_research` etc. are not available, NewsCatcher
is not set up: see **Setup** below. Install only if the human asked you to; otherwise tell them how and stop.

## Pick the tool

```
Latest news / what happened / coverage of X (last 30 days)  → search_news
How many articles / every article / exact keywords           → list_articles (count_only=true for just the number)
Every matching article into a file (CSV/JSON, up to 1,000)   → list_articles(save_as="csv")
Who covers a topic: top outlets, authors, countries          → list_articles(tally=["source","country","language"]) (one pass)
News older than 30 days (up to 12 months)                    → list_articles with from/to
Volume over time, spikes, share of voice                     → count_articles
How big is one story, who reported it first                  → get_story (story id from search_news)
A table of events with sources (deals, layoffs, recalls…)    → start_research(query) → get_research
Recent news about specific companies                         → start_research(companies) → get_research
Keep a research table or company watch updated               → start_research(keep_updated) or create_monitor
Reusable company sets (portfolio, accounts, competitors)     → save_company_list / get_company_list / edit_company_list
Credits, limits, plan                                         → get_account
```

Not a fit: general web pages, documentation, how-to advice, static company facts (founding year, HQ, investors):
use web search. To read an article itself, open its URL with a web fetch tool; NewsCatcher gives headlines,
one-line summaries, sources, dates and links.

## Costs (credits, from one balance)

| Call | Cost |
|---|---|
| search_news | 1 (low) / 5 (medium, default) / 10 (high) |
| list_articles | 5 per page (save_as: 5 per 100 articles) |
| count_articles | 5 per bucket per query (max 60 calls) |
| get_story | 1 |
| get_account, get_research, get_company_list, get_monitor, get_monitor_updates | free |
| start_research, monitors | billed per run (fast < standard < max; company runs per company); failed and empty runs are free |

Confirm with the user before standard/max runs, runs over 20 companies, hourly monitors or more than 30 count calls.

## Before a paid run (start_research, create_monitor)

- **Ask once, only for what's missing.** If the time window or the size is missing, ask a single multiple-choice
  question (the host's question picker when it has one), recommended option first, e.g. window: "Last 30 days
  (Recommended)", "Last 7 days", "Since a date"; size: "Fast, 50 rows (Recommended)", "Standard, verified with a
  briefing", "Max". Put both in the same question round; never ask twice in one run.
- **Skip the question** when the user gave both, or said "just run it": then fast, 50 rows, last 30 days.
- **Big jobs** (standard/max, more than 20 companies, keep_updated): call start_research with the same arguments
  plus `dry_run=true` first (free) and show its time and cost in that one question.
- **Monitors only on request.** Offer keep_updated or create_monitor when the user asks for something regularly
  ("every week", "keep me posted", "alert me", "track"). Don't push one on a one-off request.
- **Describe events, not coverage.** "Layoffs at US tech companies in October 2026, with company, jobs cut and
  date", not "news articles about tech layoffs". Event + who/where + window + the fields you need.

## While a run is going

- Say one line when it starts ("Running, usually 3–10 minutes; I'll post the table when it's done.") and nothing
  more until the results, unless the user asks.
- Keep calling `get_research(run_id, wait_seconds=300)` until it finishes; it waits on the server, so no sleep
  timers are needed. Don't build the table from other searches meanwhile.
- In a chat app that does not bring you back later, do not end your turn to wait: keep calling get_research in this
  turn. If you are running low on steps, stop cleanly: say it is still running, give the run id, and ask the user
  to say "refresh".
- The run keeps going if you stop; "refresh" later means `get_research` on the same run id, never a new run.

## Habits that give better answers

- search_news: start with effort=medium; use group=story when you want distinct events, min_newsrooms to keep only
  stories several newsrooms covered, window (`2h`, `1d`, `7d`, `30d`) for recency.
- list_articles is exact boolean matching: quote names (`"Goldman Sachs"`), OR synonyms (`(layoff OR layoffs OR
  "job cuts")`), search_in=["title"] for precision. When paging keep `to` fixed to the first page's window end.
- Cite every claim with the article link; say when coverage is thin (1 newsroom) or one-sided.
- Research runs take minutes: start_research, then get_research(run_id, wait_seconds=120–300). Show at most 25
  rows; its "Counts over all N rows" line gives totals (types, actors, investors, amounts) without paging.
  export_research writes the full table to a file; `exclude_rows` leaves out duplicates you found.
- Reports: search_news, list_articles and count_articles take `save_as` (local setups), and the pr-coverage-report,
  share-of-voice, competitor-intelligence and weekly-digest skills turn those files into an HTML report with
  `scripts/render-report.mjs`, so numbers and clips are never retyped.

## Setup

- **Claude Code (plugin):** `/plugin marketplace add <repo>` then `/plugin install newscatcher@newscatcher`; it asks
  for the API key once and keeps it in secure storage.
- **Claude Code (MCP only):** `claude mcp add --transport stdio --env NEWSCATCHER_API_KEY=$NEWSCATCHER_API_KEY newscatcher -- npx -y newscatcher-mcp`
- **Cursor / Claude Desktop / Codex:** an MCP server entry running `npx -y newscatcher-mcp` with env
  `NEWSCATCHER_API_KEY`.
- API keys: https://platform.newscatcherapi.com. Never paste a key into a file the user commits.

Check it works: call `get_account` (free). A plan name and balance means the key works.

## When something fails

| Error | Do |
|---|---|
| 401 | Key missing or wrong: set NEWSCATCHER_API_KEY / reconfigure the plugin |
| 402 | Balance too low: tell the user to top up; nothing was charged |
| 403 | Plan limit (limit/page_size too big, Enterprise field, mode or rows): lower it |
| 422 | Fix the parameter the message names; don't retry unchanged |
| 429 | Wait the seconds given, retry once |
| 503 on list_articles with an old window | The archive part can be down; keep to the last 30 days and say so |
| n/a buckets in count_articles | Those buckets could not be counted (archive outage or timeout); report the others and name the gap |
| Run still going | get_research again with a longer wait_seconds |

Never invent results, ids or links when a call fails: report the failure and what you tried.
Files in this package (2)
  • SKILL.md 7 KB
  • scripts/render-report.mjs 45 KB

Versions

The database has no other skills named newscatcher.

Database snapshot: 13.11.2026 · package version cf06e538f234 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 90event-dataset
    Use when the user wants a table, list, spreadsheet or dataset of real-world events from the news — \"all X since Y\", deals, incidents, announcements, appointments, recalls, lawsuits, openings, closures, protests, contracts, any repeated event type — with one row per event and a…
  • 100news-alerts-integration
    Use when the user wants to build news alerts into their own code or tools: send NewsCatcher monitor updates or new matching articles to Slack, Teams, email, a webhook, a database or a ticket queue on a schedule. Shows working code against the NewsCatcher Agent and News REST…
  • 86news-coverage
    Find current news coverage of a topic, company, person or event for a given period, with dated, cited headlines and the articles themselves. Use for the latest news, what happened this week, news roundups and reading a linked news article in full, even without naming a tool.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.