excrtx-news-sales-ai

Automates research, curation, and publication of sector news to Sales-AI via Supabase.

1|Updated May 26, 2026
One-click install
npx skills add https://github.com/elderbernardi/exocortex.saas --skill excrtx-news-sales-ai-elderbernardi
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: excrtx-news-sales-ai
Source: https://github.com/elderbernardi/exocortex.saas/tree/main/skills/excrtx-news-sales-ai
Command: npx skills add https://github.com/elderbernardi/exocortex.saas --skill excrtx-news-sales-ai-elderbernardi

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? It removes the manual work of monitoring Brazilian retail sector news, curating relevant signals, and publishing them to the Sales-AI platform, while preventing duplicate or retired news items from being republished. ## Core Features & Use Cases - Autonomous macro pipeline: A cron-driven dispatcher checks per-area cadence from a TOML config, runs research, builds a deterministic dossier, curates candidates, and publishes through the canonical Sales-AI MCP writer. - Read-before-write guard: Candidates are partitioned against existing noticias_publicas rows so active URLs are skipped and retired URLs are never reactivated. - Micro news matching: Deterministic signal-to-client matching uses frozen seller rosters with brand aliases, plus a single low-cost LLM self-gate for editorial decisions. - Use Case: Schedule a weekly run that researches the retail and cleaning sectors, curates up to 4 fresh items per area, and publishes them to the Sales-AI news feed without manual intervention. ## Quick Start Ask the agent to run the news pipeline in autonomous mode for the varejo area and publish the curated items to Sales-AI.

Frequently Asked Questions about excrtx-news-sales-ai

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate news publishing to Sales-AI?▼

Configure monitored areas in config/noticias.toml, then run the autonomous mode: news_dispatch.py identifies due areas, research scripts collect signals, build_dossier.py creates a dossier, and curated items publish via the Sales-AI MCP publish_noticia tool.

How does the skill prevent duplicate or retired news from being republished?▼

A read-before-write guard in news_guard.py partitions candidates against existing noticias_publicas rows by normalized URL and cliente_id. Active URLs are skipped, retired URLs are never reactivated, and the publish_noticia writer provides a second layer with skipped_retired results.

Does the news pipeline require DataBrain to be running?▼

No. The default autonomous mode publishes without DataBrain running and without DocBrain. DataBrain acts only as an on-demand harness for contracts, context, guard, ledger, and receipts in the optional advanced runbook.

What credentials are needed for the Sales-AI news publisher?▼

It uses a least-privilege Supabase principal with SUPABASE_URL, SUPABASE_ANON_KEY, and password-mode auth via SUPABASE_USER_EMAIL and SUPABASE_USER_PASSWORD. The service role key is explicitly forbidden in the publication path.

How are news signals matched to specific clients for micro news?▼

signal_match_prepare.py scores signals against frozen seller rosters using CNPJ, brand aliases with title-prefix rules, fantasy names, and city/UF evidence. Exact matches proceed to a single low-cost LLM self-gate that decides publish, review, or discard.

When should I not use this news pipeline?▼

Do not use it to publish without context or evidence, to run DataBrain as an active worker process, or to push RSS/LLM collection into DataBrain. Local documents from DocBrain supplement context but never substitute for a real public news URL.