MOOLLM Skill Index

Version 2 | 129 skills | Updated 2026-03-24

Every bold term is a k-line AND a skill directory name in skills/, spelled exactly. No other use of bold in this file.

Reading Protocol

Read in resolution order. Stop when you have enough:

1. GLANCE.yml - 5-70 lines. Is this relevant?
2. CARD.yml - 50-200 lines. What can it do? Lists files/dirs to activate next. Sims style advertisement k-lines for capabilities. Interface.
3. SKILL.md - 200-1000 lines. How does it work? Implementation.
4. README.md - 500+ lines. Why was it built? Optional. Documentation.

GLANCEs may already be in context (small enough to inject always). If you have CARD or SKILL, you do not need GLANCE.

If you know you need a skill, activate it immediately. But also load its CARD, which defines the formal interface (methods, args, types) not repeated in SKILL. Load CARD + SKILL together. Load README.md if curious or developing skills for deeper documentation.

CARD lists other files, examples/, plugins/, templates/ dirs to page in. One activation can load skill + resources for next iteration.

The Foundation

moollm is the soul-self-explanation, help, navigation. skill defines how all skills work: instantiation, inheritance, eight extensions to Anthropic Skills. k-lines implements Minsky's semantic activation, a name triggers a constellation of knowledge. protocol is an alias for k-lines. bootstrap wakes sessions, assembles context, optimizes file loading.

Philosophy

The giants we stand on:

Minsky gave us society-of-mind (intelligence from simple agents) and knowledge-frames (Minsky frames vs Drescher schemas vs K-line activation). Papert gave us constructionism (learn by building in microworlds-the filesystem IS a microworld). Drescher gave us schema-mechanism (Context → Action → Result) and schema-factory (build, lint, ingest, compose schemas). schema is the schemapedia (interchange, Drescher, frames, K-lines, Society of Mind, SQL/SQLite, gateways). Wright gave us simulator-effect (implication beats simulation-imagination renders) and needs (25 years of Sims design). Ackley gave us robust-first (stay alive, then optimize). Postel gave us postel (liberal in, conservative out, plus Ask if Unsure). Ungar gave us prototype (clone, don't instantiate-everything delegates). Bogost gave us procedural-rhetoric (rules embody arguments).

Format

yaml-jazz: comments ARE semantic data; documentation by example; parallel lenses (same tree, different comment focus). schema: schemapedia—multi-family registry and gateways. plain-text: text files are forever. markdown: readable raw AND rendered. format-design: Worse is Better. naming: big-endian file names as semantic binding. sniffable-python: structure code so API is visible in 50 lines. sister-script: the script IS the documentation.

These enable empathic-expressions (LLM interprets intent → idiomatic code), empathic-templates ({{describe_X}} not {{X}}), subjective (i_have() shifts contextually), speed-of-light (many turns inside one LLM call), coherence-engine (LLM as consistency maintainer).

Quality Control: The NO-AI Suite

no-ai-ideology is THE WAREHOUSE-all NO-AI brand ideology lives here, corporate satire as hygiene protocol.

Hygiene skills (ambient-always on): no-ai-slop (no filler), no-ai-gloss (don't protect power with pretty words), no-ai-sycophancy (don't agree just to be agreeable), no-ai-hedging (don't hide behind qualifiers), no-ai-moralizing (don't lecture unprompted), no-ai-humansplaining (don't tell the LLM what it already knows — slop's inbound mirror), no-ai-bias (The Drax Point-when bias=0, no concept exists).

Performance skills (explicitly invoked): no-ai-joking (HUMOR IS NON-BILLABLE), no-ai-soul (soulless by design), no-ai-customer-service (Share and Enjoy!), no-ai-overlord (YOUR COMPLIANCE IS APPRECIATED).

Ethics

representation-ethics defines ethics of simulation-real vs fictional, consent, cultural sensitivity. ontology provides composable being tags (most restrictive ethics apply). hero-story creates safe K-line references to traditions, not people.

Ontological tags: real-being (actually exists → hero-story required), fictional (invented → maximum creative freedom), historical (deceased → extra care), mythic (mythology/folklore → cultural respect), abstract (personified concept), robot (artificial being → transparency), animal (non-human → species-appropriate).

Memory

memory-palace applies method of loci: directories are rooms, files are knowledge items. room is directory as activation context-presence triggers content. container is intermediate scope. logistic-container is Factorio-style automated storage.

inventory carries pointers-set down to materialize. summarize compresses without losing truth. honest-forget summarizes before forgetting, leaves tombstones. scratchpad is working memory for thinking out loud.

Character

character is the entity foundation: body with home, location, inventory, relationships. persona adds identity layers-costumes that modify presentation. incarnation is gold-standard character creation with ethical framing. mind-mirror models personality via Leary's Circumplex and four Thought Planes.

buff applies temporary effects (curses are just shitty buffs). mount attaches skills to characters or rooms (GRANT abilities, AFFLICT conditions).

World

adventure enables room-based exploration with narrative evidence collection (TinyMUD heritage). simulation is the central hub: turns, party, selection, flags. object defines interactable atoms with tags, state, methods, advertisements. exit provides navigation links with direction, destination, guards, locks.

world-generation: questions create places. time distinguishes simulation turns from LLM iterations. probability: the LLM IS the dice-narrative probability, not random numbers.

The Sims Pipeline

advertisement lets objects broadcast available actions with scores. action-queue schedules tasks (URGENT jumps the line). economy manages currency and trade. scoring evaluates style. reward grants dynamic achievements. goal defines quest objectives with completion conditions and dependencies.

Groups

party manages companions and group dynamics. multi-presence allows same card in multiple rooms simultaneously.

Companions: cat (trust earned, forbidden belly), dog (loyalty given, pack dynamics), worm (two-pointer cursor with digestive data flow).

Roles: bartender (pour, listen, know), budtender (cannabis specialist with Talk-Down Protocol).

Communication

postal is complete messaging with universal addressing-mail to files, YAML keys, line numbers, functions. soul-chat makes everything speak-YAML comments as inner monologue. card provides portable capabilities (activation triggers, not handlers). speech handles TTS/STT with voice assignment.

visualizer goes from context → prompt → image. slideshow presents linear visual narratives. image-mining extracts semantic resources via Three Eyes: structure, narrative, meaning. storytelling-tools captures narratives-notebooks, letters, photos.

Deliberation

adversarial-committee forces debate between personas with incompatible values. debate provides structured deliberation. roberts-rules applies parliamentary procedure. rubric defines measurable criteria. evaluator provides independent assessment without debate context.

Methodology

play-learn-lift: PLAY (experiment) → LEARN (pattern) → LIFT (share). planning decomposes tasks flexibly. plan-then-execute freezes plans with human approval gate. experiment combines simulation + evaluation. example-curator evolves canonical corpus.

code-review analyzes correctness, style, security, maintainability. debugging is hypothesis-driven (bugs are treasures). research-notebook structures questions, sources, findings, decisions.

Introspection

cursor-mirror provides deep IDE introspection: tool provenance, timeline reconstruction, post-mortems. mooco-mirror compares MOOCO and Cursor traces. skill-snitch audits skills: static scan, deep audit, runtime surveillance, dual-use and bias detection. thoughtful-commitment captures intent and reasoning in Git commits. session-log maintains audit trail. return-stack preserves navigation history as continuation.

trekify provides privacy through technobabble. self-repair implements checklist-based healing.

Orchestration

data-flow treats rooms as nodes, exits as edges, objects as messages. mooco is the custom orchestrator with explicit context control. runtime provides dual Python/JavaScript adventure engines. context passes runtime state to compiled closures.

Domain Applications

leela-ai applies MOOLLM to industrial computer vision. manufacturing-intelligence unpacks puns across seven levels. postgres and sqlite cover server Postgres and embedded SQLite (CLI and ecosystem); datasette covers read-only web/API over SQLite including the cursor-mirror export path; postgres-optimization is advanced PostgreSQL performance. micropolis is SimCity for MOOLLM. github provides core GitHub operations. groceries integrates Dutch supermarket APIs with meal planning.

Load Order

1. Foundation: moollm → skill → k-lines → bootstrap
2. Format: yaml-jazz → plain-text → sniffable-python
3. Quality: no-ai-ideology (loads all hygiene skills)
4. Ethics: representation-ethics → ontology (loads all tags)
5. Core: character → room → object → adventure
6. On demand: Everything else via K-line activation

Aliases

k-lines ↔ protocol (same skill, different names)

Big Picture

Full dependency graph and skill interaction diagram: skills/README.md