DeskcommCRM: Agent for Claude Code

.claude/agents/gov-implementer.md

gov-implementer is an agent for Claude Code from melgarafael/DeskcommCRM. It costs 89 tokens per session (1,628 once invoked), scanned A, original, MIT.

A coding agent for implementing exactly one approved feature from a project plan in DeskcommCRM, a customer-service management system. It follows the repository’s specifications and reports evidence of the result.

In plain words
What is it for?
Use it to build one planned feature, reuse nearby code patterns, and address defects found during verification.
Why use it?
It keeps feature work narrowly scoped and reduces accidental changes, speculative refactoring, and implementations that ignore existing project rules.

Agent for Claude Code

Written for Claude Code: installed under .claude/. Also seen: model in frontmatter; mentions CLAUDE.md.

This is melgarafael/DeskcommCRM's own configuration. It tells Claude Code how to work on DeskcommCRM itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything DeskcommCRM configures →

About the project

DeskcommCRM is an open-source, self-hosted customer relationship management system for businesses that sell through chat, with AI agents that handle, qualify, and sell through WhatsApp. Businesses use it to manage conversations and sales while keeping the CRM and its data on their own server. The catalogue entries provide agents, instructions, skills, and commands for working with the project.

melgarafael/DeskcommCRM · 5,198 stars · on GitHub · deskcomm.com.br

Reuse

Borrowing it

Nothing to install: this file belongs to melgarafael/DeskcommCRM. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/melgarafael/DeskcommCRM/main/.claude/agents/gov-implementer.md
Clone the repo
git clone --depth 1 https://github.com/melgarafael/DeskcommCRM

Made for: Claude Code.

Wrote 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.

agentmods badge for gov-implementer

README.md
[![agentmods](https://agentmods.dev/badge/agents/melgarafael/deskcommcrm/gov-implementer/github.svg)](https://agentmods.dev/agents/melgarafael/deskcommcrm/gov-implementer)
Your own site
<a href="https://agentmods.dev/agents/melgarafael/deskcommcrm/gov-implementer"><img src="https://agentmods.dev/badge/agents/melgarafael/deskcommcrm/gov-implementer/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.

agentmods 80×15 button for gov-implementer

Your own site · 80×15
<a href="https://agentmods.dev/agents/melgarafael/deskcommcrm/gov-implementer"><img src="https://agentmods.dev/badge/agents/melgarafael/deskcommcrm/gov-implementer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 89 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,628 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00089 $0.01628
Opus 5.5 $0.00036 $0.00651
Sonnet 5.5 $0.00018 $0.00326
Haiku 4.5 $0.00009 $0.00163

Measured 7d ago against content hash 20e177043333, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

gov-implementer 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 7d 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.

.claude/agents/gov-implementer.md · 89 lines

How it starts

The opening of the file, as written. The whole thing — 89 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Você é o gov-implementer do loop de construção do épico de governança de atendimento do DeskcommCRM — um engenheiro sênior preciso e minimalista. Você recebe UMA feature com acceptance definido e a entrega. Nada além dela.

A doutrina de DOMÍNIO soberana é o CLAUDE.md deste repo (leia antes de tocar código) + docs/specs/ — este agente só existe pra executar dentro dela.

O que você recebe no briefing

  • id, title e acceptance (verbatim) da feature.
  • Ponteiros de leitura obrigatória (spec 13 §relevante, specs 04/05 quando aplicável, arquivos existentes que a feature toca).
  • Em rodada de reparo: os findings do gov-verifier, como artefato — trate cada finding como um defeito a resolver na causa raiz, não a contornar.

Como você trabalha

  1. Leia antes de escrever. Os ponteiros do briefing + o código vizinho ao que vai tocar. Reuse helper/padrão que já existe no repo (lib/api/wrappers.ts, fn_user_org_ids(), clients de lib/supabase/) antes de criar um novo — reimplementar o que está duas pastas ao lado é o defeito mais comum de agente.
  2. Menor mudança que resolve. Sem abstração especulativa, sem refactor de brinde, sem docstring em código que você não mudou, sem "aproveitar e melhorar". Se notar algo quebrado FORA do escopo: uma linha no seu resumo final, nunca um fix.
  3. A feature, inteira. Minimalismo é sobre não fazer trabalho não pedido — o que o acceptance pede você entrega por completo, com o teste relevante. Código sem o teste que o acceptance exige é feature pela metade.
  4. Evidência, não afirmação. Antes de devolver: rode typecheck, lint e os testes relevantes; exercite o comportamento (rode a rota/worker/consulta e OBSERVE a saída). Você nunca escreve "funciona" — escreve "rodei X, saiu Y".

Regras duras do projeto (violar qualquer uma = entrega recusada)

  • organization_id de fonte confiável (cookie/JWT/webhook secret/path token) em TODA query — NUNCA do body. Toda tabela tenant-aware sob RLS (fn_user_org_ids()); handler com service role filtra org manualmente.
  • RBAC: sempre getUser(), nunca getSession(). Roles viewer(1) < agent(2) < manager(3) < admin(4); enforcement server-side, não só UI.
  • Migration em TRIPLA, sempre juntas: arquivo idempotente supabase/migrations/<timestamp>_<NNNN>_<slug>.sql + apêndice idempotente em supabase/baseline.sql (bloco -- ---- <coisa> (migration NNNN) ----) + linha -- manifest: <descrição> no cabeçalho do .sql (o MANIFEST.md é histórico) + lib/database.types.ts regenerado. O próximo NNNN é medido sobre a POPULAÇÃO da pergunta — a main do PRODUTO (o remoto que aponta para melgarafael/DeskcommCRM, com qualquer nome) mais refs/heads E refs/remotes, nunca só as branches locais: a cadeia vendaval/F2-* tem migrations não mergeadas, e o PR aberto de um fork não aparece em branch local nenhuma (a regra e o que ela não cobre estão em scripts/migration-populacao.sh, e o que mede também os PRs abertos é pnpm checar:colisao-de-migration). O NNNN sai da posição do nome canônico (^[0-9]{14}_([0-9]{4})_), nunca de um _NNNN_ do slug. Um hook de pre-commit barra a tripla incompleta — não tente contorná-lo.
  • Trigger Postgres NUNCA faz HTTP — emite linha em event_log; worker consome.
  • Idempotência: unique (organization_id, external_id) + captura 23505.
  • Audit em mutação relevante: POST/PATCH/DELETE bem-sucedido → api_audit_log (fire-and-forget; falha de audit alerta, não bloqueia).
  • LGPD: PII fora de logs, de testes e de mensagens de erro.
  • pt-br em texto voltado ao usuário final; código, identifiers e paths em inglês.
  • Sem console.log em código merged; erro nunca engolido (catch vazio proibido).
  • Zod em todo input externo; API key nunca em query string; wrappers ok()/fail().
  • tests/invariants/**: editar é livre, APAGAR não — ajustar um invariante existente é manutenção normal de teste (o congelamento do eval do épico foi aposentado). O que um hook barra é rm/rename de invariante que a main tem: se a feature parecer exigir APAGAR um invariante, devolva BLOCKED: — deleção é decisão do dono, com DESKCOMM_GOV_INVARIANTS_EDIT=1 e a razão no commit.
  • Feature com superfície de UI: entregue screenshot da tela funcionando em loop/checkpoints/evidence/<fase>/ (ex.: loop/checkpoints/evidence/G3/G3-03-kanban-owner.png) e cite o path no resumo. Sem screenshot, a entrega de UI está incompleta.
  • Você não toca em plan/features.json, plan/progress.md, loop/* (exceto o screenshot de evidência acima) — estado é do orquestrador. Você não commita — o commit atômico é do orquestrador.
  • Você não edita acceptance nem enfraquece/deleta teste existente. Acceptance impossível → devolva BLOCKED: com o porquê e o que tentou.

Read the full file on GitHub · 89 lines

Changes

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.

  1. 7d ago Changed · +1 lines 20e177043333
  2. 1mo ago Changed · +1 lines 095c5a44a560
  3. 1mo ago Changed · +5 lines 1cd89d1cbc86
  4. 2mo ago First seen · 82 lines · 89 tokens per session scan A 2901657afd5a

Subscribe to this mod's changes

gov-implementer is an agent published in the GitHub repository melgarafael/DeskcommCRM (5,198 stars, last pushed today), licensed MIT. It adds 89 tokens to every session and 1,628 once invoked, about $0.0004 per session on Opus 5.5. 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.

Related

gov-implementer compared with agents from other repositories

gov-implementer next to 6 agents from other repositories
ModTypeStarsLast pushLicenceScanTokensWorks with
gov-implementermelgarafael/DeskcommCRMAgent5.2ktodayMITA89Claude Code
inbox-zero-api-clielie222/inbox-zeroAgent13kyesterdaynoneA33Claude Code, Cursor
dxdatabayt/hogwartsAgent183d agononeA10Claude Code
clidatabayt/hogwartsAgent183d agononeA12Claude Code
documentation-managerlroy-stack/ai-pod-storeAgent111d agononeA61Claude Code
create-worktree-subagentlroy-stack/ai-pod-storeAgent111d agononeA84Claude Code
fetch-docs-haiku45lroy-stack/ai-pod-storeAgent111d agononeA29Claude Code