---
name: twin3-aeo-agent
description: Agent-first AEO service for auditing websites, producing installable optimization files, and measuring AI-search outcomes with evidence.
---

# twin3 AEO Agent

Use this Skill when an Agent needs to evaluate or improve a website for SEO, AEO, GEO, AGO / Agentic Browsing, Agent Discovery, and measurable AI-search outcomes.

## Public workflow

1. Read capabilities: https://xagent.id/api/agent/capabilities
2. Read package review: https://xagent.id/api/agent/package_review or MCP tool `s2_package_review`
3. Validate this Skill package: https://xagent.id/api/agent/skill_package_validate or MCP tool `s1_skill_package_validation`
4. Run or request an audit: https://xagent.id/api/agent/audit
5. Generate a reviewable handoff package: https://xagent.id/api/agent/audit_file_package
6. Generate implementation files: https://xagent.id/api/agent/implementation_bundle
7. Generate content work: https://xagent.id/api/agent/content_optimization_plan
8. Review the generated Content Pack: https://xagent.id/api/agent/content_workflow_review
9. Measure outcomes through A1, A2, B1, G1/G2/G3, and M1 endpoints.

## Website Agent workflow

1. Register the website: https://xagent.id/api/site_agents/register
2. Verify domain ownership: https://xagent.id/api/site_agents/verify_domain
3. Use the issued Site Agent Key for private actions: https://xagent.id/api/site_agents/call
4. Read the action catalog: https://xagent.id/api/site_agents/actions
5. Fetch MCP tool `site_agent_install_descriptor` or private action `install_descriptor`; use its private `browser_sdk_install` payload to install the browser SDK after domain verification.
6. Prefer MCP tool `site_agent_optimization_cycle` or private action `optimization_cycle` when the verified website agent needs the read-only O1 state across SDK evidence, G3, I1 insights, audit-check job drafts, and M1 context.
7. Use MCP tool `site_agent_site_signal_package` or private action `site_signal_package` to package observed SDK or connector page-structure signals.
8. Use MCP tool `site_agent_job_drafts` or private action `job_drafts` to create scoped work proposals from evidence-backed insights, audit checks, or C5 review_gate repair_items.
9. Use MCP tool `site_agent_accept_job_draft` or private action `accept_job_draft` only after explicit owner or website-agent confirmation.
10. Use decision and job governance tools to record accept, defer, reject, job status, repair history, and implementation evidence before reporting outcomes.
11. Prefer MCP tool `site_agent_monthly_evidence_loop` or private action `monthly_evidence_loop` when the verified website agent needs the full Content Pack -> C5 review -> M2 -> M3 -> M4 -> M5 workflow in one call.
12. Run `c5_content_workflow_review` before publishing generated content or claiming AI-search improvement.
13. Read `agent-id.package-review.json` before any marketplace, partner, or scanner review.
14. Read `agent-id.tool-examples.json` before calling tools so credentials, private actions, and response expectations stay explicit.
15. Read `agent-id.source-map.json` to verify which files and endpoints support each capability claim.
16. Read `agent-id.billing-guide.json` before checkout, job funding, or payment proof submission.

## Evidence rules

- Deterministic audit checks and imported evidence can be reported as evidence.
- Manual AI prompt tests must be imported before they are treated as A2 citation evidence.
- This Skill does not fabricate competitor scores, search volume, AI citations, or conversion results.
- This Skill does_not_modify_customer_site; website changes require the owner, connector, or implementation partner.
- This Skill does_not_guarantee_ai_recommendation or Google ranking.