Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/nett0eth/research-stack/setup-research-stacknpx skills add nett0eth/research-stack --skill setup-research-stackgit clone --depth 1 https://github.com/nett0eth/research-stackWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00074 | $0.00991 |
| Opus 5 | $0.00037 | $0.00495 |
| Sonnet 5 | $0.00015 | $0.00198 |
| Haiku 4.5 | $0.00007 | $0.00099 |
Grade A, and why
setup-research-stack scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Setup do research-stack
Instale as ferramentas, registre o tempo real de cada etapa e relate o que ficou faltando. Não silencie falha: ferramenta ausente vira aviso explícito.
Etapa 0: descobrir onde o Bash roda
Antes de qualquer instalação, verifique se o Bash disponível é a máquina do usuário ou um ambiente isolado.
Rode ls / e confirme se o sistema de arquivos do usuário está acessível. No Claude Code em terminal local, está. No Cowork e em ambientes com sandbox, o Bash é outro computador e instalar ali não serve para nada, porque os servidores MCP rodam na máquina do usuário.
Se o ambiente for isolado, não rode o script. Entregue os comandos para o usuário rodar no terminal dele e siga para a Etapa 2 no mesmo formato. Diga isso explicitamente em vez de instalar no lugar errado e reportar sucesso.
Etapa 1: rodar o script
Execute ${CLAUDE_PLUGIN_ROOT}/scripts/setup.sh com Bash. Ele instala os conversores e os clientes MCP em Python, cronometra cada instalação e grava o resultado em ./research-stack-timings.txt.
Peça confirmação ao usuário antes de rodar, porque o script instala pacotes na máquina dele. Ofereça a flag --dry-run para ele ver a lista antes.
Etapa 2: skills de terceiros
Este plugin não redistribui skill de outro autor. Instale da fonte original, preservando crédito e mantendo tudo atualizado.
Caminho padrão, sem custo e sem chave de API. Rode um por vez e reporte quais funcionaram:
npx skills add mattpocock/skills --skill research --agent claude-code
npx skills add tw93/waza --skill learn --agent claude-code
npx skills add 199-biotechnologies/claude-deep-research-skill --skill deep-research --agent claude-code
npx skills add lllllllama/RigorPilot-Skills --skill ai-research-explore --agent claude-code
npx skills add coreyhaines31/marketingskills --skill competitor-profiling --agent claude-code
Essas cinco usam as ferramentas nativas do agente e as bases públicas do paper-search, então funcionam sem credencial paga.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 73 lines · 74 tokens per session scan A 1997cf906135
setup-research-stack is a skill published in the GitHub repository nett0eth/research-stack (4 stars, last pushed 1mo ago), licensed MIT. It adds 74 tokens to every session and 991 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
arxiv-fetch
Fetch the latest arXiv papers and generate an on-demand digest. Use when the user asks to fetch/check today's (or the latest) arXiv papers, find new papers on a topic, or generate an ad-hoc arXiv digest. The agent judges relevance itself — no API keys needed.
scholar-sidekick
This skill should be used when the user mentions a scholarly identifier (DOI, PMID, PMCID, ISBN, arXiv, ISSN, NASA ADS bibcode, WHO IRIS URL) and wants structured metadata, a formatted citation, a bibliography export file, a retraction check, an open-access check, or verification that a claimed citation is real (not…
read-research-paper
Renders ANY research paper (URL / arXiv ID / DOI / PDF / pasted text) into a visually engaging multi-format reading experience — with mind maps, method flowcharts, key-finding infographics, comparison tables, related-work timelines, and a plain-English layer alongside the technical content. Activates on slash commands…
research-paper
Enterprise-grade autonomous research paper generation skill for AI coding agents — full papers, literature reviews, theses, whitepapers, surveys, policy briefs — with rigorous methodology, statistical validation, multi-style citations (Harvard / APA / IEEE / MLA / Chicago / Nature / arXiv-numeric), and rich…
get-research-paper
Discovers, retrieves, ranks, and summarizes real existing research papers on any topic. Searches arXiv, Google Scholar, PubMed, Semantic Scholar, and reputable open repositories; returns a curated reading list with verified DOIs, key findings, and citation-ready metadata. Activates on slash commands…
ideogram-ultra
Build Ideogram 4 (Ideogram Ultra) txt2img and img2img workflows with the local open-weights model, dual conditional/unconditional models with DualModelGuider, Qwen3-VL text encoder, and structured JSON ("compositional deconstruction") prompts for strong text rendering and layout control.