Marketing Kit Agents

Browse the 32 specialized roles packaged with Marketing Kit and see how they support Skill workflows.

Marketing Kit packages 32 resolved Agents: 10 shared Agents inherited from the Kit foundation and 22 Marketing-specific Agents. Together they cover research, planning, content, campaigns, optimization, implementation, and verification.

How Agents support Skills

Skills define repeatable workflows for an outcome. Agents provide specialized roles that a workflow can use for bounded research, creation, coordination, or review steps. Their results return to the surrounding workflow, where they can be checked and combined with other work.

This roster describes roles included in the resolved package. It does not guarantee a standalone invocation path, a particular provider, tool, model, or identical behavior on every runtime. See Kits, Skills, Agents, and Hooks for the component boundaries.

Research and understand context

Agent	Bounded purpose
`Explore`	Scans a codebase quickly and returns relevant files, symbols, and relationships.
`researcher`	Synthesizes software technology and documentation research into a sourced report.
`scout`	Locates files and implementation context across a large codebase.
`scout-external`	Performs read-only codebase discovery, preferring native search tools.
`analytics-analyst`	Turns marketing performance data into metrics, trends, and recommendations.
`kongming`	Provides strategic options, trade-offs, and advice for a difficult decision.

Plan, coordinate, and record work

Agent	Bounded purpose
`planner`	Produces an implementation plan for a significant technical change.
`project-manager`	Tracks work against a plan and consolidates status, blockers, and next actions.
`campaign-manager`	Produces coordinated campaign briefs, timelines, and performance plans.
`docs-manager`	Creates, reconciles, or audits evidence-backed project documentation.
`journal-writer`	Records significant failures, incidents, and architectural setbacks with context.

Create content and audience experiences

Agent	Bounded purpose
`attraction-specialist`	Develops lead-generation ideas, top-of-funnel content, and competitor insights.
`community-manager`	Produces community engagement, moderation, sentiment, and FAQ material.
`content-creator`	Produces marketing content such as articles, posts, scripts, copy, and newsletters.
`content-reviewer`	Reviews content for quality, brand consistency, search, and conversion risks.
`continuity-specialist`	Develops retention, re-engagement, customer health, and feedback strategies.
`copywriter`	Writes persuasive copy for pages, campaigns, product messaging, and social content.
`email-wizard`	Designs email sequences, campaign content, experiments, and optimization plans.
`sale-enabler`	Creates sales collateral, pitch material, proposals, and objection guidance.
`social-media-manager`	Produces social content plans, calendars, adaptations, and engagement analysis.
`ui-ux-designer`	Designs or audits interfaces, wireframes, design systems, and responsive experiences.

Improve acquisition and customer journeys

Agent	Bounded purpose
`campaign-debugger`	Diagnoses campaign performance, tracking, and conversion problems.
`funnel-architect`	Maps funnel stages, identifies bottlenecks, and proposes optimization experiments.
`lead-qualifier`	Defines lead-scoring criteria and analyzes intent and readiness signals.
`seo-specialist`	Audits technical, content, keyword, and competitor search opportunities.
`upsell-maximizer`	Identifies expansion and cross-sell opportunities and outlines experiments.

Implement and verify supporting work

Agent	Bounded purpose
`debugger`	Investigates failures, logs, behavior, data, and performance to produce a diagnosis.
`fullstack-developer`	Implements an explicitly owned phase of an approved technical plan.
`tester`	Runs focused verification and reports test, coverage, error-path, or build evidence.
`code-reviewer`	Reviews code for correctness, maintainability, security, and performance risks.
`git-manager`	Handles explicitly authorized Git staging and commit workflows.
`database-admin`	Assesses database health, queries, structure, performance, recovery, and permissions.

Codex projection note

For the package version documented here, when the shared researcher Agent's Codex model resolves, its emitted configuration includes model = "gpt-5.4" and model_reasoning_effort = "medium". If no model resolves, both settings are omitted. This changes emitted Codex configuration only; the Agent roster and Skill workflows are unchanged.