메뉴
Use this skill when the user needs to search NBER working papers, inspect NBER paper metadata, download NBER paper PDFs, run NBER-CLI commands, or configure NBER-CLI as an MCP server.
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/SepineTam/nber-cli --skill nber-cli명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
SKILL.md 표시 중
| name | NBER-CLI |
| description | Use this skill when the user needs to search NBER working papers, inspect NBER paper metadata, download NBER paper PDFs, run NBER-CLI commands, or configure NBER-CLI as an MCP server. |
Use this skill when the user wants to use NBER-CLI for NBER working papers, including:
nber-cli is already installed.uvx nber-cli ... for one-off usage because it can run the published package without requiring a checked-out project or preinstalled command.nber-cli ... is also fine.Check the command:
uvx nber-cli --help
uvx nber-cli --version
Search working papers:
uvx nber-cli search "labor economics"
uvx nber-cli search "minimum wage" --start-date 2024-01-01 --end-date 2024-12-31
uvx nber-cli search "ChatGPT" --format json
Show paper metadata:
uvx nber-cli info w25000
uvx nber-cli info w25000 --all
uvx nber-cli info w25000 --format json
Download papers:
uvx nber-cli download w34567
uvx nber-cli download w34567 --save-base ~/papers/nber
uvx nber-cli download w34567 --file ~/papers/nber/w34567.pdf
uvx nber-cli download --batch w34567 w25000 w32000 --save-base ~/papers/nber
Run the MCP server:
uvx nber-cli mcp-server
uv run nber-cli mcp-server --transport streamable-http --port 8000 --yes
The first command uses the published package's default stdio server and includes the download tool. The second uses the current source tree's Unreleased transport spelling and exposes only search and paper-info tools; --host defaults to 127.0.0.1, and the Docker image explicitly uses --host 0.0.0.0 so Compose can forward its port. --port already exists in published 0.4.0, while the source tree newly requires --yes for a custom value. The HTTP transport has no built-in authentication; do not expose it to an untrusted network.
The examples/ directory contains focused guides for advanced workflows:
examples/use-of-feed.md: incremental RSS feed monitoring, JSON filtering, cron automation, systemd timers, and batch downloads.examples/use-of-search.md: historical catalog queries, date ranges, pagination, and search-to-download chains.examples/use-of-info.md: metadata lookup, full records, batch inspection, and ID validation.examples/use-of-download.md: single and batch downloads, save paths, restricted access handling, and feed-to-download integration.examples/use-of-mcp-server.md: stdio and HTTP MCP server setup, client configuration, and security notes.examples/use-of-json-pipeline.md: stable JSON shapes, jq recipes, citation extraction, and append-only JSON Lines logs.examples/use-of-db.md: database initialization, custom paths or sqlite:///... URLs, migration, direct SQLite queries, and backup.examples/use-of-automation.md: weekly digest scripts, cron recipes, systemd timers, and topic alerts.examples/use-of-error-handling.md: interpreting 403, 404, and timeouts, plus robust shell patterns.The feed subsystem tracks NBER's new working papers RSS feed (https://www.nber.org/rss/new.xml) in the local database. It is designed for incremental monitoring, not exhaustive search. Use it to answer questions like "what papers appeared this week?" or "has any new paper on topic X shown up?"
For more complete examples, including JSON filtering, cron automation, systemd timers, and batch-download scripts, see examples/use-of-feed.md.
The database and tables are created automatically on the first feed fetch, but you can also initialize explicitly:
uvx nber-cli db init
This writes the default path ~/.nber-cli/nber.db to ~/.nber-cli/config.json. To keep feed data in a custom location:
uvx nber-cli db init --db-path ~/research/nber.db
uvx nber-cli db init --db-path sqlite:////Users/name/research/nber.db
uvx nber-cli db migrate ~/Dropbox/research/nber.db
db migrate moves the database file and its SQLite sidecar files to the new path and updates config. The database stays on the user's machine and is accessed through SQLModel/SQLAlchemy.
feed fetch always downloads the current RSS feed and updates last_seen_at for every item. By default it only prints items that were not already in the cache:
uvx nber-cli feed fetch
After the first run, the same command becomes an incremental "what's new" check. To see everything that NBER is currently advertising, regardless of cache state:
uvx nber-cli feed fetch --display-all
To preview a fixed number of the most recent items (and automatically enable --display-all):
uvx nber-cli feed fetch --max-items 10
uvx nber-cli feed fetch --max-items 10 --format json
Use JSON output when you want to filter, store, or pass feed items to another tool:
uvx nber-cli feed fetch --format json | jq '.items[] | {paper_id, title, authors}'
Save only new papers to a JSON Lines file for downstream processing:
uvx nber-cli feed fetch --format json | jq -c '.items[]' >> nber_feed_new.jsonl
The JSON shape is stable and includes:
source_url: the RSS URLdatabase_path: resolved path to the local SQLite cachetotal_fetched: how many items NBER returnednew_count: how many items were not already in the cachedisplay_all: whether the output includes all items or only new onesmax_items: the display cap, if anyitems: list of feed items, each with paper_id, title, authors, abstract, url, source_url, guidA typical agent workflow is:
feed fetch to discover new papers.search if you need a broader or historical set.info to read abstracts and metadata.download to retrieve PDFs.Example: fetch the latest feed, then inspect and download a specific paper:
uvx nber-cli feed fetch --max-items 5
uvx nber-cli info w32000 --all
uvx nber-cli download w32000 --save-base ./papers/this-week
Example: find the newest paper by an author you follow:
uvx nber-cli feed fetch --display-all --format json \
| jq '.items[] | select(.authors[] | contains("Daron Acemoglu")) | {paper_id, title}'
Then download the matching paper:
uvx nber-cli download w32000 --save-base ./papers/acemoglu
feed clean removes rows from feed_items. It always previews the match count and asks for confirmation unless you pipe y to it. The audit table feed_fetches is never touched, so you keep a complete history of every fetch.
Remove items not seen in the last 30 days (the default):
uvx nber-cli feed clean
Use a custom retention window:
uvx nber-cli feed clean --days 7
Clean a specific date range:
uvx nber-cli feed clean --start-date 2026-05-01 --end-date 2026-05-31
Wipe the cache entirely (all feed_items rows):
uvx nber-cli feed clean --all
This is safe because the next feed fetch will repopulate whatever NBER is currently advertising. It is useful when you want to reset the "new" state, for example after changing how you filter papers.
Because feed fetch is idempotent and fast, you can run it on a schedule. A minimal cron entry that fetches every weekday morning and appends new items to a JSON Lines log:
0 9 * * 1-5 cd /home/user/research && uvx nber-cli feed fetch --format json | jq -c '.items[]' >> nber_feed_new.jsonl
A slightly richer script that downloads new papers to a weekly folder:
#!/usr/bin/env bash
# save_new_papers.sh
WEEK_DIR="./papers/$(date +%Y-%W)"
mkdir -p "$WEEK_DIR"
uvx nber-cli feed fetch --format json > /tmp/feed_new.json
if [ "$(jq '.new_count' /tmp/feed_new.json)" -gt 0 ]; then
jq -r '.items[].paper_id' /tmp/feed_new.json | while read -r paper_id; do
uv run nber-cli download "$paper_id" --save-base "$WEEK_DIR" --restrict true
done
fi
nber-cli search ... for that.new.xml RSS feed.nber-cli download for that.feed clean to manage size.Use default text output for humans. Use JSON when the result should be consumed by another program or agent:
uvx nber-cli search "inflation" --format json
uvx nber-cli info w25000 --format json
Interpret download errors conservatively:
403 means NBER denied access for the current IP, institution, account, session, download limit, or paper access policy.404 means the paper or PDF endpoint was not found.NBER-CLI must not bypass NBER controls. Do not suggest proxy rotation, credential sharing, CAPTCHA bypass, account automation, request-signature tampering, or other access-circumvention behavior. If NBER returns denial, limits, redirects, or access pages, surface the response clearly and stop.
If the user finds a bug or unexpected behavior in NBER-CLI, direct them to report it at https://github.com/sepinetam/nber-cli/issues.