alphaxiv

name	alphaxiv
description	Use when the user wants to work with the alphaXiv CLI to search for papers, inspect paper metadata or full text, manage saved paper or assistant context, or use the authenticated assistant, folders, or comments in terminal-based research workflows.

alphaXiv CLI

Use this skill when the user wants to use the alphaXiv CLI to map topic vocabulary, find papers, inspect one paper, build a shortlist, validate blueprint claims against anchor papers, or use the authenticated assistant from the terminal.

This skill is CLI-only. Do not switch to Python SDK guidance unless the user explicitly asks for Python usage instead of terminal usage.

Mental model

• search: keyword lookup when the user already has terms in mind and wants matching papers, topic suggestions, or organizations.
• feed: recent or ranked discovery from the public alphaXiv homepage feed. Use this for questions like "important recent papers" or when the user needs topic slugs before filtering.
• paper: inspect one paper. This is where to get metadata, the original abstract, AI summary, long overview, extracted paper text, PDF access, similar papers, comments, and paper-specific folder membership.
• assistant: authenticated synthesis after deterministic retrieval. Use this to explain, compare, or continue a research chat after search, feed, or paper.
• context: save or inspect the current paper or assistant session so later commands can omit ids.
• auth: manage API-key auth and optional browser-backed web login for assistant commands.

Highest-confusion distinctions

• search vs feed: use search for keyword-driven lookup; use feed for recent or ranked discovery and for topic-filter workflows.
• paper abstract: the author-written abstract.
• paper summary: the short AI digest.
• paper overview: the longer AI write-up.
• paper text: readable text extracted from the PDF.
• paper pdf download: the actual PDF file saved locally.
• assistant start vs assistant reply: start opens a new assistant chat; reply continues the current or selected saved session.
• Long-running assistant chats may get slower over time; if that happens, use assistant start for a fresh session and carry forward only the minimal context you still need. This is the available mitigation when the slowdown is coming from session growth.

Auth rules

Public flows:

• search
• feed
• most read-only paper commands such as show, abstract, summary, overview, overview-status, resources, text, similar, pdf, and comments list
• context

API-key-authenticated flows:

• folders
• paper vote
• paper view
• paper folders
• paper comments add
• paper comments reply
• paper comments upvote
• paper comments delete

Assistant-authenticated flows:

• all assistant commands

Assistant commands can authenticate through either:

• alphaxiv auth set-api-key
• alphaxiv auth login-web

Prefer auth login-web when assistant chat writes are restricted for the API key. Treat auth login-web as one-time setup for a persistent browser profile, not something to rerun before every assistant command.

Working rules

1. Prefer deterministic retrieval first: search, feed, and paper.
2. Use assistant after retrieval when the task becomes comparison, synthesis, or follow-up reasoning.
3. Use context use paper <paper-id> when several paper commands will target the same paper.
4. Do not run mutating commands unless the user explicitly wants to change state.
5. If a new assistant chat should be grounded in one paper, pass --paper <paper-id> to assistant start. A saved current paper does not automatically ground a new assistant chat.
6. If a user reports a slow long-running assistant session, explain that assistant reply keeps extending the same remote session. Suggest a fresh assistant start plus a short recap or the relevant paper ids, while keeping the older session available through assistant history or assistant list. Frame this as a mitigation for session-growth slowdown, not a guaranteed fix for every latency issue.
7. When the task is blueprint or draft validation rather than paper discovery, stay narrow: reopen the exact anchor papers, inspect them with deterministic commands, and use the assistant only to pressure-test whether the current wording or synthesis is defensible.
8. In validation mode, do not reopen broad search unless the existing anchor set clearly fails to support the active claims.
9. In article-development work, this validation mode is often the highest-value use of alphaXiv. Do not default to broad discovery when the real problem is an overclaimed blueprint.
10. When an assistant session materially changes a package thesis, vocabulary, claim boundary, paper queue, or workflow-design decision, track it in the package's assistant-discussion index and create a compact extraction note. Use the workflow guidance in ../../../workflows/alphaxiv-research/README.md rather than pasting raw transcripts into the package.

Reference map

Read only the reference file you need:

• ../../../workflows/alphaxiv-research/README.md: tracked canonical research workflow for discovery, validation, shortlist building, and assistant use.
• ../../../workflows/topic-discovery/README.md: pre-package editorial workflow for turning a broad idea or research seam into a topic map, candidate slate, and package-init or parking decision.
• references/workflows.md: task workflows that combine commands into research tasks.
• references/command-map.md: concise command-family map, auth requirements, mutation notes, and nearby-command disambiguation.
• references/rollout-lessons.md: browser-auth persistence, assistant overload mitigation, validation-mode lessons, and feed-filter discovery patterns recovered from real usage.