Skip to content

Using Kairos through an agent

Kairos ships instructions an AI agent can follow to run it for you: set it up, find and rank jobs, explain a match, change what scores weigh, add searches, and report what scoring costs. You still decide and apply; the agent sends nothing on your behalf without your approval.

Claude Code

Install the skill once, then ask in plain words ("find me backend jobs in Melbourne", "why did this one score 72?", "weigh work-life balance higher"):

mkdir -p ~/.claude/skills
cp -r skills/kairos ~/.claude/skills/

Inside this repository Claude Code also reads AGENTS.md and CLAUDE.md, which point to the same skill.

Other agents

Point the agent at AGENTS.md in the repository root, or paste skills/kairos/SKILL.md into its instructions. Everything it needs runs through the kairos CLI, so any agent with a shell can follow it.

What the agent needs from you

  • The setup in Onboarding done, or permission to do it.
  • Your resume as a Markdown file, and answers to its one batch of questions (where you would work, what matters to you, whether you need visa sponsorship).
  • Credentials in the environment, never in chat: CLAUDE_CODE_OAUTH_TOKEN or an LLM API key, and optionally Adzuna's keys.

The skill is checked against the CLI in CI (tests/test_cli/test_agent_skill.py), so a renamed command fails the build instead of leaving an agent with stale instructions.