Show HN: AgentKit SEO - a career context file that installs as agent skills

3 points

a month ago

I was job hunting across countries and kept running into the same two problems.

First: every platform optimises for different things. LinkedIn wants keywords. ATS parsers want specific formatting. GitHub rewards README quality. A portfolio has to be crawlable. Staying consistent across all of them without a system is nearly impossible.

Second: every AI agent session starts from scratch. Paste a context dump, watch it burn half the window fetching your links, get generic output. The agent has no grounded facts about you – so it improvises.

Both problems point to the same fix: one verified source of truth, written once, readable by humans and agents alike.

That's AgentKit SEO [1]. The core concept is an agent-context file – a private Markdown document with your career facts, project metrics, and proof links. You build it once using the context skill. Every other skill (LinkedIn, GitHub, CV/ATS, portfolio, X) reads from it.

It ships as a CLI that installs focused skill files into whichever agent you use:

  npx agentkit-seo install --provider codex
  # or: claude-code, gemini-cli, opencode, or portable SKILL.md

This is not an auto-apply tool. It's the step before that: create one grounded file, then use it as input for any agent, any platform, any task. No hallucination from context overload. No contradictions between your CV and your LinkedIn.

Open source, free, MIT licence. My co-author Elia Innocenti and I built it because it didn't exist.

[1] https://github.com/agentkit-seo/agentkit-seo

[2] https://agentkit-seo.github.io/

[3] https://www.npmjs.com/package/agentkit-seo