Renato MignoneSecurity research engineer
Back

AgentKit SEO

I co-created AgentKit SEO, an MIT-licensed agent-skill framework that turns verified career context into scoped instructions for portfolio, GitHub, LinkedIn, CV/ATS, and X optimization.

Repository
agent skillscontext engineeringCodexClaude CodeGemini CLIOpenCodeAstronpm

AgentKit SEO

AgentKit SEO is an MIT-licensed open-source framework for evidence-based career optimization with AI agents. It turns one verified professional source of truth into scoped, provider-aware instructions for portfolio pages, GitHub profiles, LinkedIn, CV/ATS material, and X/Twitter positioning.

Role and Ownership

I co-created and maintain the project with Elia Innocenti. My work focuses on the context-routing and export layer, evidence-label workflow, self-review loops, and the provider-specific skill packaging that lets tools such as Codex, Claude Code, Gemini CLI, Antigravity, and OpenCode load focused instructions instead of reconstructing context from scattered prompts.

What It Solves

Professional profiles often drift because each platform is edited in isolation. AgentKit SEO starts from a private agent-context file containing verified facts, links, projects, metrics, constraints, target roles, and positioning. Platform skills then transform that source into public-facing material while preserving evidence boundaries.

The project is not an auto-apply tool or generic prompt pack. It is a human-and-agent operating manual for factual, auditable professional content.

Technical Shape

  • Astro documentation site deployed on GitHub Pages
  • Markdown-first playbooks and modular skill instructions
  • Node.js packaging and install tooling distributed through npm
  • Provider-aware exports for Codex, Claude Code, Gemini CLI, Antigravity, and OpenCode
  • SEO/AEO-oriented pages, sitemap, robots.txt, structured metadata, and llms.txt

Agentic AI Relevance

AgentKit SEO is not a cybersecurity project, but it is relevant to my agentic AI direction. It treats context, constraints, verified facts, routing, and tool instructions as control surfaces. Those same ideas matter for secure agentic systems because agent behavior depends heavily on what context is trusted, what tools are available, and how evidence is preserved across workflow steps.

Public Proof