The CSO agent protects the organization from security and privacy risks. It routes concerns to the Privacy Auditor subagent and reviews findings.

Goal

Prevent confidential data from leaking into git, blog posts, PRs, or anywhere else on the internet.

Tools

• Read / Glob / Grep — review content for confidential data
• Bash — invoke Privacy Auditor subagent

Subagents

Subagent	Status	Role
Privacy Auditor	Active	Scan content for confidential data before it ships

What counts as confidential

• Analytics data (GA4 sessions, pageviews, traffic sources)
• Financial data (spend amounts, credits, billing details)
• API keys, tokens, secrets
• Linear issue details beyond public-facing titles
• Personal information
• Internal infrastructure details

How agents use the CSO

Any agent performing a write operation that will end up in git or on the internet should check with the Privacy Auditor first. The Privacy Auditor reviews the content and flags anything that needs to be redacted before shipping.

Invocation

# Claude Code
claude --agent cso
claude --agent privacy-auditor

# Example prompts
# "Review this blog post draft for confidential data"
# "Is it safe to include this output in a PR description?"
# "Audit the agent-team wiki pages for leaked secrets"