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.
git clone --depth 1 https://github.com/georgemarmelstein/superjurista-devWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/georgemarmelstein/superjurista-dev/pipeline-pesquisa)<a href="https://agentmods.dev/commands/georgemarmelstein/superjurista-dev/pipeline-pesquisa"><img src="https://agentmods.dev/badge/commands/georgemarmelstein/superjurista-dev/pipeline-pesquisa/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/georgemarmelstein/superjurista-dev/pipeline-pesquisa"><img src="https://agentmods.dev/badge/commands/georgemarmelstein/superjurista-dev/pipeline-pesquisa.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00040 | $0.06247 |
| Opus 5 | $0.00020 | $0.03123 |
| Sonnet 5 | $0.00008 | $0.01249 |
| Haiku 4.5 | $0.00004 | $0.00625 |
Grade A, and why
pipeline-pesquisa scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 271 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Orquestrador: Pipeline de Pesquisa de Precedentes v3.0
v3.0 — retomada + gate por script + fontes verbatim (molde vivo:
.claude/commands/pipeline-sentenca.md). O que mudou da versão anterior: (1) RETOMADA — o workspace é DETERMINÍSTICO (slug do tema, SEM timestamp), então pesquisa cujo artefato já existe e passa no gate não roda de novo; (2) validação DETERMINÍSTICA —scripts/verificar_pesquisa.pyconfere as âncoras (normalizadas de acento/caixa); o orquestrador NÃO lê relatórios para validar; (3) FONTES VERBATIM — cada pesquisador grava também um parcialfontes-<fonte>.jsonescripts/merge_fontes.pyproduz o corpus$ID-fontes.json(cadeia de custódia das citações, consumida porscripts/verificar_citacoes.py); (4) subagente responde UMA LINHA de status — o relatório vive no arquivo, nunca na conversa; (5) FONTES COMPLETAS — além de BNP/CJF/JULIA, o pipeline pesquisa STJ (espelhos SCON via Dados Abertos) e TNU (base viva do eproc), fechando a lista de fontes autorizadas. Os métodos de busca (sintaxe de cada MCP) vivem nos agentes pesquisadores.
<scripts_deterministicos> | Script | Função | |--------|--------| | scripts/verificar_pesquisa.py | Gate + retomada: varredura (PENDENTES), --etapa (exit-coded), --etapas (subconjunto), --gate (final) | | scripts/merge_fontes.py | Funde os parciais fontes-*.json em $ID-fontes.json (valida, deduplica, reatribui IDs); sem LLM, sem contexto | </scripts_deterministicos>
<agents_utilizados> | Agent | Capacidade | Arquivo | |-------|------------|---------| | pesquisador-bnp | Precedentes vinculantes STF/STJ | .claude/agents/pesquisa/pesquisador-bnp.md | | pesquisador-cjf | Jurisprudência regional TRF1/TRF3/TRF4 (radar de bases vivas) | .claude/agents/pesquisa/pesquisador-cjf.md | | pesquisador-julia | Jurisprudência TRF5 (por turma) | .claude/agents/pesquisa/pesquisador-julia.md | | pesquisador-stj | Jurisprudência STJ (repetitivos, súmulas, dominante) | .claude/agents/pesquisa/pesquisador-stj.md | | pesquisador-tnu | Jurisprudência TNU (representativos, uniformização JEFs) | .claude/agents/pesquisa/pesquisador-tnu.md | | consolidador-pesquisa | Análise cruzada e hierarquia | .claude/agents/pesquisa/consolidador-pesquisa.md | </agents_utilizados>
<regras_uso>
- RETOMADA: antes de despachar, o gate diz o que já está válido — o que está OK não roda de novo. Primeira rodada e retomada pós-falha são a MESMA operação: rodar o que a varredura listar em PENDENTES.
- CONDUZIR POR CAMINHO: o orquestrador passa paths prontos; o subagente pesquisa via MCP e GRAVA (Write) o relatório E o parcial de fontes no workspace. O documento NUNCA volta inline na resposta.
- RESPOSTA DE UMA LINHA: cada subagente responde apenas " OK | $ID-pesquisa-.md" — quem confere o conteúdo é o script, não o orquestrador lendo.
- VALIDAÇÃO POR SCRIPT: nunca validar lendo o documento; sempre python scripts/verificar_pesquisa.py "$WORKSPACE" --etapa <fonte>.
- Subagentes LEEM o próprio prompt via Read (.claude/agents/pesquisa/...); o orquestrador não copia a capacidade deles — injeta só o TEMA, os caminhos e o lembrete de sintaxe da fonte.
- As pesquisas 1a/1b/1c/1d/1e são INDEPENDENTES entre si: despachar as pendentes em PARALELO (até 5 Tasks sonnet no MESMO turno). A consolidação (Etapa 2) só roda depois delas.
- Subagentes nunca usam TodoWrite.
</regras_uso>
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.
- 11d ago First seen · 271 lines · 40 tokens per session scan A 2a6c0458a86c
pipeline-pesquisa is a command published in the GitHub repository georgemarmelstein/superjurista-dev (22 stars, last pushed 1mo ago), licensed MIT. It adds 40 tokens to every session and 6,247 once invoked, about $0.0002 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 2026-08-30.
Other commands, from other repositories
deps-audit
Audit project dependencies for vulnerabilities, outdated packages, license conflicts, and supply chain risks — then provide actionable remediation strategies.
review-epo-claims
Analyze patent claims for EPO Art. 84 EPC compliance - clarity, conciseness, support by description.
voice-compliance
Voice/telephony compliance check — invokes voice-ai-reviewer to produce TM-voice-{slug}.md with TCPA, STIR/SHAKEN, state recording-consent, EU AI Act Art. 50, and synth-voice deepfake-law gaps.
review-policy
Review a policy document for structure, coverage, and language quality.
at-dsgvo
You are helping an enterprise architect generate an Austrian Data Protection Assessment — the Austrian-specific GDPR layer applied by the Datenschutzbehörde (DSB) under the Datenschutzgesetz (DSG 2018, BGBl. I Nr. 165/1999 as amended). Run this after /arckit:eu-rgpd to add Austrian obligations that go beyond the EU…
au-pspf
You are an enterprise architect generating a Protective Security Policy Framework (PSPF) compliance assessment for an Australian Government entity or contractor handling government information.