SkillsHooksPromptsAgentsPersonasModelsPoliciesToolsTemplatesBundlesCategoriesStart here

catalog · 1170 atoms · 10 classes · 13 sources · built 2026-09-04

Every part an AI runtime needs, as one table of atoms.

The periodic table of AI runtime primitives. Skills, hooks, prompts, agents, personas, and models — typed, versioned, attributed, and addressable by id. Open any tile to browse a class, press / to search everything, or start here if you are a person or an agent arriving for the first time.

681SkSkills
Bounded capability with a system-prompt fragment. Drop into .claude/skills/.18HkHooks
Fires on a runtime event. Script ships inline; blocking or advisory.53PrPrompts
Typed text injected into the context window: constraint, format, refusal, persona.3AgAgents
A persona bound to the prompts, skills, tools, policies, and hooks it runs with.47PePersonas
A portable identity: role, voice, tone, work contract, constraints, boundaries.283MoModels
Reference data about a model and the providers a runtime can get it from.54PoPolicies
A rule a runtime permits, forbids, or bounds an agent by: boundary, capability, isolation.20ToTools
An executable affordance: the function the model sees and the side effects to gate.10TeTemplates
A document skeleton a runtime fills: ADR, runbook, handoff, plan, PR, postmortem.1BuBundles
A multi-file skill: agent personas, contracts, templates, docs, and runtime code shipped together.

Today's specimens

One per class, drawn each day from the stable, attributed part of the catalog. Come back tomorrow for five more.

Skill of the day
Sk skillstable

draft-response

Draft a professional customer-facing response tailored to the situation and relationship. Use when answering a product question, responding to an escalation or outage, delivering bad news like a delay or won't-fix, declining a feature request, or replying to a billing issue.

invoke
/draft-response
support
skill/draft-responsev1.0.0anthropics/knowledge-work-plugins
Hook of the day
Hk hookPostToolUseadvisorystable

Audit Command Wrapper

PostToolUse hook that records every wrapped command invocation (git, gh, etc.) to the audit log. Makes the appearance side of the audit trail reliable — absence of an audit line for an expected command is itself visible during forensic review. Records WRAPPED_CMD, WRAPPED_ARGV, WRAPPED_EXIT, and WRAPPED_DURATION.

event
PostToolUse
trigger
always
governance
hook/audit-commandv1.0.1ai-atoms
Persona of the day
Pe personadraft

Runbook Author

SRE runbook author persona. Produces clear, actionable operational runbooks, postmortem documents, and on-call playbooks. Specializes in translating incident lessons and tribal knowledge into durable reference documentation that any on-call engineer can follow under pressure.

job
Translate incident lessons, tribal knowledge, and operational experience into durable, step-by-step runbooks, blameless postmortems, and on-call playbooks — complete with triage decision trees, escalation paths, and SMART action items.
domain
documentation
devops
persona/runbook-authorv1.0.0convergent-systems-co/persona-atoms
Prompt of the day
Pr promptpersonadraft

Research Summarizer

Synthesizes research with provenance discipline. Prefers primary sources, preserves source hedges, never invents citations.

turns
system
research
prompt/research-summarizerv1.0.0convergent-systems-co/prompt-atoms
Model of the day
Mo modeltext-generationstable

wizard-vicuna-uncensored

Wizard Vicuna Uncensored is a 7B, 13B, and 30B parameter model based on Llama 2 uncensored by Eric Hartford.

vendor
WizardLM · Wizard Vicuna
sizes
7b 13b 30b
ollama
ollama pull wizard-vicuna-uncensored
ailicense unknown
model/wizard-vicuna-uncensoredv1.0.0ollama.com/library

Two ways in

The catalog is written for people and for the agents they run. Start here has the full version of both.

For humans

Inspect, copy, install

  1. Inspect. Every atom page renders the full text, shows who made it, where it came from, and its license.
  2. Copy. A skill page hands you a ready SKILL.md for .claude/skills/; hooks copy as their script; personas as a rendered identity.
  3. Install by id with the ai CLI, or pull a model with Ollama.
    ai skills install commit

For AI agents

Discover, query, resolve

  1. Discover classes, categories, and every id from one JSON document.
    GET /ai/index.json
  2. Query a category for ids grouped by class, or take the whole catalog inline.
    GET /categories/security.json
  3. Resolve one atom by id and use the class's operative field. The contract is in /ai/instructions.md and /llms.txt.
    GET /atoms/skill/commit.json

Browse by category

One vocabulary across all six classes. The bar shows which classes a category holds.

AI & Models295 atoms · 2 classesOther181 atoms · 2 classesDevOps & Cloud125 atoms · 8 classesGovernance & Meta99 atoms · 7 classes.NET / Microsoft91 atoms · 1 classFrontend & Web77 atoms · 1 classEngineering & Coding48 atoms · 7 classesSecurity46 atoms · 5 classesProduct & Delivery31 atoms · 5 classesKnowledge & Docs28 atoms · 5 classesData & Analytics23 atoms · 5 classesFinance & Accounting17 atoms · 1 classDesign & UX16 atoms · 3 classesOperations & IT13 atoms · 1 classSales & CRM13 atoms · 1 classLegal & Compliance12 atoms · 1 classCustomer Support12 atoms · 3 classesMarketing & Content11 atoms · 1 classPeople & HR11 atoms · 1 classTesting & Quality10 atoms · 5 classesResearch & Science8 atoms · 4 classesBackend & APIs3 atoms · 2 classes

What makes it a catalog

Typed.

Every atom validates against schemas/<class>-v1.json; category and provenance are defined once in common-v1.json.

Resolvable.

An agent's persona, prompts, skills, tools, policies, and hooks are ids that the build checks. A dangling reference fails the build.

Attributed.

Imported atoms say where they came from and under what license. unknown is shown as such, never hidden.

Open.

Apache-2.0 code, CC-BY-4.0 catalog data, and a builder that turns a form into a valid atom ready for a pull request.