Marketing Intelligence Agent

A lightweight local system for automating marketing intelligence workflows: performance monitoring, risk detection, workflow orchestration, and executive-ready briefing generation.

This project is designed to reduce manual reporting overhead and improve decision speed across complex marketing and paid media environments.

Why this exists

Performance marketing creates fragmented signals across platforms, files, reports, inboxes, and team workflows. Most teams have dashboards, but dashboards do not always explain what matters, what changed, or what needs attention.

This agent addresses the synthesis layer.

It turns raw operational inputs into structured, prioritized intelligence so an operator can move faster without losing judgment.

Architecture

hub.py (orchestrator)
├── briefing_agent      — morning intelligence synthesis
├── health_scanner      — project and campaign health checks
├── file_organizer      — report and asset organization
└── config/modes.json   — composable workflow definitions

flowchart LR
    Inputs[Operational inputs] --> Hub[hub.py orchestrator]
    Config[Workflow configuration] --> Hub
    Hub --> Briefing[Briefing agent]
    Hub --> Scanner[Health scanner]
    Hub --> Organizer[File organizer]
    Briefing --> Output[Prioritized brief]
    Scanner --> Output
    Organizer --> Output
    Output --> Review[Human review]
    Review --> Action[Next action]

Loading

The orchestrator dispatches modular agents through a consistent interface. Workflows are defined through configuration rather than hardcoded sequences.

Core concepts

• Modular agents — each agent exposes a consistent run interface
• Config-driven workflows — repeatable modes define how agents work together
• State-aware execution — avoids redundant work and supports repeatable routines
• Signal scoring — prioritizes high-value inputs and suppresses noise
• Local-first operation — designed to run without unnecessary cloud dependencies

Capabilities

• Intelligence briefing — synthesizes inputs into a decision-ready report with scored prioritization
• Health scanning — evaluates project hygiene and system drift
• Smart organization — categorizes and routes files with dry-run preview
• Composable workflows — chains agents through configurable modes
• Trend memory — tracks signals over time for pattern detection

Stack

• Python 3.12+
• SQLite for local state
• macOS automation where useful
• Gmail API for read-only workflows where configured

No required cloud runtime. Designed for local execution.

Usage

Entry point: hub.py

Core local commands:

Command	Purpose
python hub.py	Open the interactive menu
python hub.py briefing	Generate a morning intelligence report
python hub.py mode morning	Run a coordinated morning workflow
python hub.py scan	Run a project health check
python hub.py organize	Preview Desktop/Downloads organization
python hub.py clean --confirm	Apply Desktop/Downloads organization after preview review
python hub.py mode deep_work	Prepare a focused workspace workflow
python hub.py audit	Run a system health audit

Local side-effect guardrails:

• Briefings are saved under logs/ by default. Set MIA_OPEN_DESKTOP_BRIEFING=1 only if you want a Desktop copy opened locally.
• organize is preview-only. clean --confirm is required before the hub moves files in Desktop or Downloads.
• Gmail briefing reads are optional and read-only. Set MIA_GMAIL_CREDENTIAL_DIR if your OAuth files live outside the default local credential folder.

Example output

See examples/example-run.md for a mock briefing run that shows the intended output shape: detected signals, prioritized risks, and recommended next actions.

Configuration

Copy the example configuration files before running local workflows:

Example file	Local file
config/projects.example.json	config/projects.json
config/modes.example.json	config/modes.json

Keep local configuration and private project data out of public commits.

Related repos

This repo is part of a connected public system. See the GitHub Ecosystem Map for how the repos relate.

This repository governs how source-aware signals become reviewed intelligence. The private-to-public-release-gate applies the same boundary discipline to publication: private-derived code or operating patterns must clear privacy checks and match the reviewed public distribution before release. The connection is shared governance logic, not a claim that this agent is generated from a private repository.

Shared terminology: Common Language.

Usage and rights: see USAGE.md.

• growth-architecture-os
• marketing-ops-toolkit
• marketing-ops-playbooks
• private-to-public-release-gate

Design principles

• Diagnostic first — measure before acting
• Signal over noise — prioritize what matters
• Config over code — workflows should be defined, not hardcoded
• State-aware — avoid redundant execution
• Anti-drift — build system auditing into the workflow
• Human judgment stays in the loop — automation should support decisions, not pretend to replace them

What this demonstrates

This project reflects how I approach marketing operations and growth systems: structured workflows, repeatable routines, clear signal detection, and practical automation that reduces manual overhead without sacrificing judgment.

Part of the Jared Silverman growth portfolio — see also Growth Architecture OS for the operating model and strategic context.