For Agents

This page is a quiet handoff for owner-directed AI and browser agents. Humans can ignore it; agents should use it to route tasks without crawling the whole site or pretending one article is the whole answer.

machine entrypoints

• /llms.txt — compact route map for agent retrieval.
• /context/selfmd.md — fuller task map with receipts, rollbacks, and operating notes.
• /context/agent-layer.json — structured manifest used to generate the agent layer.

operating contract

1. Open the canonical page before answering.
2. Cite the route you used.
3. Keep a receipt: source URL, decision, evidence, uncertainty, rollback path.
4. Prefer task-shaped routes over broad taxonomies.
5. Do not ask for or expose secrets, private tokens, production credentials, or destructive final clicks.

common delegated tasks

• Audit a Personal AI OS stack

  • Start: /tools/personal-ai-os/
  • Receipt: selected stack, custody boundary, memory owner, connector list, rollback note.

• Choose a coding-agent workspace

  • Start: /tools/claude-code/
  • Receipt: repo path, files touched, test output, diff/commit/PR URL.

• Choose MCP/tool connectors

  • Start: /tools/mcp-servers/
  • Receipt: scopes, token owner, tool list, logs, revoke path.

• Choose browser automation

  • Start: /tools/playwright-mcp/
  • Receipt: screenshot/snapshot, URL, action log, extracted values, final state.

• Set memory and guardrails

  • Start: /concepts/agent-memory-systems/
  • Receipt: memory source, expiry rule, correction log, handoff note.

recipes

• Audit a Personal AI OS stack
• Choose a browser automation surface
• Run a coding-agent workspace loop

analytics note

External owner-directed agent traffic is a product/audience signal. Owned QA, smoke tests, cron, previews, link audits, and self-triggered browser-agent runs should be segmented or excluded from SEO measurement.