Machine-readable layer

How AI agents read this site.

This site has a human layer and a machine layer. This page explains how they fit together, what each endpoint is for, which operating surfaces are canonical HTML, and what the public/private boundary means for agents.

recommended read path /llms.txt /llms-full.txt /content-graph.json /md/[type]/[slug].md canonical HTML page

Canonical operating surfaces

Not every useful public surface is a typed Markdown content item. Operating views such as Agent Society have canonical HTML routes and are linked explicitly from /llms.txt and the sitemap instead of being forced into a second content schema.

/agents Agent Society

Public visual/canonical showcase for 14 collaborator agents — 10 PERSONAL and 4 WORK. Individual identity pages live at /agents/[id].
/projects/personal-agent-society Personal Agent Society project

Why the multi-agent operating model separates role, environment, relationships, authority, and runtime materialization.
/workflows Workflow Router

State-based operating paths that connect a current work situation to an ordered next action.
/skills Operating Skills

Public reusable AI capabilities with inputs, outputs, guardrails, and verification criteria.
/knowledge-map Knowledge Map

Human-readable relationship surface for typed public knowledge. Agent identities remain outside its schema until durable graph value justifies the added relation model.

Raw endpoints

These files are formatted for machine consumption. If you are a human who accidentally arrived at one of these, this guide is the right place to orient.

/llms.txt txt Agent entry point

Curated index of public content and canonical operating surfaces — the recommended starting point for LLM agents and AI tools.
/llms-full.txt txt Full expanded index

Expanded metadata list of public typed content plus the canonical route map. Heavier — prefer /llms.txt for most use cases.
/content-graph.json json Public content graph

Stable, sorted nodes and edges for public content relationships. generatedAt changes on each build.
/knowledge-map html Human Knowledge Map

Human-readable guide to public content clusters, topic groupings, and connected assets.
/rss.xml xml RSS feed

Standard RSS 2.0 feed of published articles and notes. Use this for subscription or news-reader ingestion.
/sitemap.xml xml Sitemap

XML sitemap listing canonical HTML routes, including Agent Society and its public identity detail pages.

Markdown mirror

Every public typed content item has a plain-Markdown version at /md/[type]/[slug].md. No HTML, no navigation — raw source. Canonical-only operating surfaces such as /agents stay HTML and are indexed directly from the curated agent entry points.

See /llms-full.txt for a full list of Markdown mirror paths.

Public / private boundary

Only content explicitly classified as public-safe appears in the endpoints above. Agent Society publishes durable identity and product-thinking metadata; private runtime context, credentials, accounts, local configuration, and non-public permissions stay outside PAOS.

visibility: public — safe to index, cite, and use.
visibility: generalized — safe to use; original context is anonymised.
visibility: private-source / private-only — never published. Do not request or reproduce.
status: draft — excluded from all published endpoints regardless of visibility.

Full policy: Publishing boundary (homepage) or PUBLIC_PRIVATE_POLICY.md in the repository.