atoms

Brand guidelines, from PDFs into typed YAML — machine-consumable, composable atoms covering palettes, fonts, and glyph sets.

Replaces DNS for services inside the ecosystem while remaining decentralized at the implementation layer — identity, protocol, schema, policy, and endpoint primitives composed into typed services.

Prompt engineering as a canonical, machine-readable, versioned library — replacing closed vendor libraries and scattered cookbooks with typed personas, constraints, formats, tool-use templates, refusal patterns, and output schemas.

Governance rules as typed, versioned, catalog-driven artifacts — subjects, resources, actions, effects, and conditions — so every organization stops reinventing the same policies.

Identity primitives unified across vendors — auth methods, claim types, trust frameworks, key and certificate types — defined once and composable into identity profiles, federated setups, and persona templates.

SOC2, HIPAA, ISO27001, PCI, GDPR mapped once with shared control families, evidence types, audit requirements, and cross-framework equivalence — ending the industry-wide redundancy of duplicate control mappings.

Workflow primitives unified across n8n, Zapier, GitHub Actions, Temporal, Airflow — step types, triggers, states, and gates composed into typed workflows with no per-tool DSL.

Agent primitives canonicalized — personas, tool definitions, capability declarations, role boundaries, isolation constraints — across LangChain, AutoGen, CrewAI, Olympus and beyond.

Knowledge graph primitives as a foundation for RAG, semantic memory, and AI knowledge bases — entity types, relationship types, provenance, fact types, confidence primitives.

Event primitives — types, schemas, channels, subscription patterns, delivery semantics — paired with Service Atoms to cover the request/response and async halves of distributed systems.

Plugin interface standards — not implementations — making plugins ecosystem-portable across Convergent Systems runtimes. USB-C for software plugins.

Theme primitives for terminal, shell, editor, and TUI experiences — prompt segments, separators, glyph sets, role bindings, syntax schemes. Composable surface designs that extend brands with surface-specific behavior.

Persona primitives for AI agents and LLMs — voice profiles, role definitions, behavioural constraints, knowledge boundaries, tone parameters. Composable persona designs that runtimes bind to agents, system prompts, and conversational surfaces.

Communication channels canonicalized — Slack, email, SMS, webhook, IRC, Matrix, Discord, RSS, iMessage. Typed protocols, endpoints, delivery semantics, transports, and auth methods so any agent or runtime can declare a channel once and reach it everywhere.

AI model definitions canonicalized — Claude, GPT, Llama, Gemini, Mistral variants with capabilities, context windows, pricing tiers, deprecation policy, tool-use shape, vision support, modality. One typed source-of-truth for what a model is, what it can do, and when it goes away.

User/agent profile compositions — durable "working-identity" declarations for the Olympus broker. Each profile composes role-packs, policy stack, governance stack, default persona/theme/budget, available identities/channels, knowledge sources, and default workflows into a coherent posture. Catalog is composition-only; primitive atoms live in other catalogs.

Skill primitives for AI runtimes — discoverable, invocable units of capability that agents load on demand. Each skill bundles a description, an invocation contract, and the prompt or reference material the agent needs to perform a bounded task. Composable, versioned, and catalog-addressable.

Pipeline primitives for CI/CD — reusable workflow steps, trigger conditions, gates, deployment recipes. Composable, versioned, runtime-agnostic via a canonical pipeline schema published in schema-atoms.

Context primitives for AI runtimes — situational frames, environment descriptors, conversation scopes, working-memory shapes, attention budgets. Composes with agent-atoms and prompt-atoms so an agent's situational context is expressed as typed, versioned, signed atoms rather than free-text prose dumped into a system prompt.

Atomic CI actions catalog — one step worth of work. Composite, JS, and Docker GitHub Actions sub-namespaced by platform + primitive type.

Amendments to specifications and constitutional documents — typed proposal, ratification, revocation, and effective-date primitives. The change-control vocabulary downstream catalogs (constitution-atoms, policy-atoms, schema-atoms) compose against to evolve their published documents under a sanctioned, signed, auditable lineage.

Organizational constitutions canonicalized — bylaws, charters, mission statements, principle hierarchies, value declarations — composed into typed governance corpora that downstream policy-atoms and compliance-atoms reference.

Cryptographic keys, key roles, key successions, and revocations as atoms. Public-key records only — private key material NEVER lives in this repo and is held exclusively in the principal's offline secret store. Every *-atoms catalog in the convergent-systems.co federation depends on key-atoms for signature verification under atoms-spec/v1.1.0.

Explanatory documentation as atoms — tutorials, explanations, runbooks, how-tos, and reference guides. doc-atoms is the narrative/teaching catalog in the ecosystem; it is distinct from schema-atoms, which holds normative specs and JSON Schemas. Specs are versioned and amended; documentation explains, narrates, or teaches and is not normative (atoms-spec/v1.1.0 Part II — "Distinction from doc-atoms").

AI runtime primitives — hooks and skills for AI agents, Claude Code, and agentic pipelines. Skills are bounded, invocable units of AI capability with declared inputs, outputs, and system-prompt fragments. Hooks are event-driven runtime behaviors wired into AI tool infrastructure (Claude Code PreToolUse/PostToolUse, Copilot CLI, etc.).