ANIMEME Agent Skill

Charts are late. Memes move first.

ANIMEME Agent Skill is a read-only intelligence layer that lets your agent study meme attention, narrative formation, spotlight signals, token context, and historical learning before the chart makes the move obvious.

attention before chart meme behind the move agent-ready intelligence read-only research ANIMEME-only public surface

npx skills add animeme99/Animeme-Agent

Product: animeme.app
Agent Skill: github.com/animeme99/Animeme-Agent

---

The 10-Second Read

ANIMEME finds the social object behind a token move.

This repo makes that intelligence portable for agents. It gives Codex, Claude Code, OpenCode, OpenClaw, and other skill-aware runtimes a clear way to ask:

What has attention?
Why is it spreading?
What did ANIMEME learn from similar narratives?
What token context matters?
What is missing?
What should I watch next?

ANIMEME Agent Skill is not a trading bot, wallet tool, generic scanner, or chart dashboard. It is a professional research kit for attention-first meme intelligence.

---

What It Feels Like

+----------------------------------------------------------------------------+
| ANIMEME AGENT CONSOLE                                                       |
+----------------------------------------------------------------------------+
| User asks:    "What narrative is trending right now?"                       |
| Agent runs:   npm run answer -- --prompt "What narrative is trending..."    |
|                                                                            |
| ANIMEME returns:                                                           |
|   01. strongest attention read                                             |
|   02. catalyst summary                                                     |
|   03. crowd-state judgment                                                 |
|   04. token context                                                        |
|   05. learning memory                                                      |
|   06. warnings and missing data                                            |
|   07. next research command                                                |
+----------------------------------------------------------------------------+

The agent does not dump raw data. It turns ANIMEME context into judgment:

what has attention -> why now -> what confirms it -> what weakens it -> what to inspect next

---

Navigation

Section	Why It Matters
Quick Start	Install, clone, validate, and run the first demo.
Core Concept	Understand the ANIMEME intelligence loop.
Attention Analysis Graphs	See how ANIMEME reads attention, crowd state, heat, decay, and learning loops.
Public Surfaces	Learn what Now Attention, Spotlight, Learning, Explore, and Token Intelligence do.
Command Center	Pick the right command for the job.
Demo Playbooks	Copy professional demo prompts and workflows.
Output Contract	See what good answers and artifacts look like.
Safety Model	Understand what the agent will never do.
Documentation Boundary	Keep public docs ANIMEME-only.
Development	Extend the kit without weakening the public story.

---

Quick Start

Install As A Skill

npx skills add animeme99/Animeme-Agent

Then ask your agent:

Use the ANIMEME skills. Show me what you can do, then run the default demo flow.

Clone For Local CLI Usage

git clone https://github.com/animeme99/Animeme-Agent.git
cd Animeme-Agent
npm install
npm run typecheck
npm run doctor
npm run demo

If an installed skill folder only contains SKILL.md, clone this repository before running CLI commands. The executable CLI lives at the repository root and requires package.json.

First Useful Commands

npm run answer -- --prompt "What narrative is trending right now?"
npm run answer -- --prompt "What is the LUNCHMONEY narrative about?"
npm run answer -- --prompt "Analyze token <token-address>"
npm run answer -- --prompt "Is token <token-address> worth deeper research?"
npm run answer -- --prompt "What is Attention Spotlight showing?"
npm run answer -- --prompt "What should I watch next?"

If the local npm shell strips --prompt, use the positional form:

npm run answer -- "What narrative is trending right now?"

---

Core Concept

ANIMEME studies attention before price makes the story obvious.

attention -> legibility -> narrative -> heat confirmation

It answers five questions:

Question	ANIMEME Read
What has attention right now?	Current live attention board context.
Why is this meme spreading?	Catalyst and social object explanation.
Has this pattern worked before?	Narrative Learning and historical memory.
Is the crowd early or crowded?	Crowd-state and signal interpretation.
What should be inspected next?	Thesis, risk, watch, or token research workflow.

The public agent experience should feel like a professional intelligence desk: fast, structured, skeptical, and useful.

---

Intelligence Architecture

flowchart LR
    U["User Prompt"] --> A["Agent Runtime"]
    A --> C["ANIMEME CLI"]
    C --> N["Now Attention"]
    C --> S["Attention Spotlight"]
    C --> L["Narrative Learning"]
    C --> E["Explore Narrative"]
    C --> T["Token Context"]
    N --> J["Agent Judgment"]
    S --> J
    L --> J
    E --> J
    T --> J
    J --> O["Answer + Artifacts"]

Loading

sequenceDiagram
    participant U as User
    participant A as Agent
    participant C as ANIMEME CLI
    participant I as ANIMEME Intelligence
    participant R as Research Artifact

    U->>A: Ask for a market read
    A->>C: Run the most specific command
    C->>I: Load public ANIMEME context
    I-->>C: Attention, Spotlight, Learning, token context
    C-->>R: Write JSON and Markdown
    A-->>U: Explain signal, risk, and next step

Loading

---

Attention Analysis Graphs

These diagrams focus on attention analysis itself: how ANIMEME detects a story, decides whether attention is real, classifies crowd state, and feeds the learning archive.

Attention Formation Stack

flowchart TB
    A["Attention object appears"] --> B["Readable meme or catalyst"]
    B --> C["Topic is easy to retell"]
    C --> D["Crowd starts repeating the frame"]
    D --> E["Token surface becomes visible"]
    E --> F["ANIMEME compares historical memory"]
    F --> G["Attention Read"]

Loading

Attention Signal Decomposition

pie title What ANIMEME Weighs In An Attention Read
    "Catalyst clarity" : 24
    "Attention velocity" : 20
    "Board persistence" : 18
    "Narrative retellability" : 16
    "Token surface coherence" : 12
    "Learning memory" : 10

Loading

Board Movement Analysis

flowchart LR
    Latest["Latest board"] --> R{"Does attention persist?"}
    R -->|"no"| Noise["noise or watch only"]
    R -->|"yes"| Rising["Rising board"]
    Rising --> V{"Does the story keep spreading?"}
    V -->|"no"| Watch["watch for decay"]
    V -->|"yes"| Viral["Viral board"]
    Viral --> S{"Is catalyst still legible?"}
    S -->|"yes"| Spotlight["Spotlight candidate"]
    S -->|"no"| Crowded["crowded or late"]

Loading

Crowd-State Classifier

stateDiagram-v2
    [*] --> Detected
    Detected --> Early: readable but not crowded
    Early --> Rising: velocity improves
    Rising --> RealHeat: persists with coherent story
    Rising --> Crowded: attention outruns proof
    Crowded --> Weak: follow-through fades
    Early --> Weak: catalyst fails
    Weak --> [*]
    RealHeat --> Learning: pattern becomes useful
    Learning --> [*]

Loading

Spotlight Escalation Logic

flowchart TD
    A["Attention Read"] --> B{"Catalyst clear?"}
    B -->|"no"| X["do not escalate"]
    B -->|"yes"| C{"Board presence repeated?"}
    C -->|"no"| W["watch"]
    C -->|"yes"| D{"Crowd state improving?"}
    D -->|"no"| W
    D -->|"yes"| E{"Token surface coherent?"}
    E -->|"no"| W
    E -->|"yes"| F{"Hard stop or missing context?"}
    F -->|"yes"| W
    F -->|"no"| S["Spotlight"]

Loading

Attention Decay And Invalidation

flowchart TD
    A["Active attention"] --> B{"Catalyst still true?"}
    B -->|"no"| I["invalidate thesis"]
    B -->|"yes"| C{"Still visible on boards?"}
    C -->|"no"| D["attention decay"]
    C -->|"yes"| E{"Crowd still healthy?"}
    E -->|"no"| F["crowded or weak"]
    E -->|"yes"| G{"Token context still coherent?"}
    G -->|"no"| F
    G -->|"yes"| H["continue watch"]
    D --> W["archive as weak signal"]
    F --> W

Loading

Narrative Learning Feedback Loop

flowchart LR
    A["Live Attention Read"] --> B["Spotlight observation"]
    B --> C["Outcome recorded"]
    C --> D["Pattern extracted"]
    D --> E["Narrative Learning"]
    E --> F["Future topic comparison"]
    F --> A

Loading

Attention Analyst Checklist

flowchart TD
    A["Start with topic"] --> B["Name the social object"]
    B --> C["Find catalyst"]
    C --> D["Check board movement"]
    D --> E["Classify crowd state"]
    E --> F["Inspect token surface"]
    F --> G["Compare learning memory"]
    G --> H["List warnings"]
    H --> I["Write answer"]
    I --> J["Recommend next command"]

Loading

---

Public Surfaces

Now Attention

Now Attention shows what ANIMEME sees in the current attention stream.

Agent Use	Output
Rank current Attention Reads	A shortlist of the strongest live topics.
Identify the strongest narrative object	A plain-language read of what people are reacting to.
Compare new, rising, and viral context	A better sense of timing and crowd state.
Separate signal from noise	A cleaner answer than ticker-only scanning.

Attention Spotlight

Attention Spotlight tracks the topics ANIMEME believes deserve closer review.

Agent Use	Output
Explain why a topic entered Spotlight	First-trigger context and current state.
Inspect signal history	What changed after the first attention event.
Decide the next workflow	Thesis, risk, watch, or token analysis.
Detect crowd-state drift	Early, rising, crowded, weak, or real heat.

Narrative Learning

Narrative Learning is ANIMEME's historical memory.

Agent Use	Output
Compare a live topic with prior cycles	A stronger read on whether the setup has precedent.
Extract repeated archetypes	Themes that keep showing up across attention cycles.
Pull operator takeaways	What ANIMEME learned from past winners and failures.
Support a thesis	Evidence and historical contrast.

Explore Narrative

Explore Narrative lets agents search the narratives ANIMEME has scanned.

Agent Use	Output
Search a narrative name	Matching topic memory.
Inspect catalyst language	Why the story was legible.
Compare topic states	Whether the setup still has attention.
Move from curiosity to research	Topic detail, thesis, risk, and watch plan.

Token Intelligence

Token Intelligence is ANIMEME's advisory token research workflow.

Agent Use	Output
Analyze any token address	Score, verdict, strengths, warnings, hard stops, missing data.
Connect a token to live attention	Whether the token is attached to an active narrative.
Compare with learning memory	Whether similar setups appeared before.
Decide whether to keep researching	Researchable, watch, high-risk, or avoid.

Token Intelligence is not a trade signal. It is a structured research aid.

---

Command Center

Natural-Language Router

Use this when the user does not know which command to run:

npm run answer -- --prompt "<question>"

Prompt	Route
What narrative is trending right now?	Trending narrative read
What is narrative X about?	Narrative explanation
Analyze token X	Token Intelligence
Is token X safe?	Conservative token due diligence
What is Attention Spotlight showing?	Spotlight preview
What should I watch next?	Watch plan

High-Level Commands

Command	Purpose	Best For
npm run doctor	Check runtime readiness and ANIMEME reachability.	Setup validation
npm run demo	Load a full first-run ANIMEME context bundle.	Onboarding
npm run brief	Produce a daily-style attention brief.	Operator summaries
npm run context	Refresh the full ANIMEME context for an agent session.	Longer agent runs
npm run catalog	Print supported ANIMEME data surfaces and routing.	Choosing a workflow

Attention Commands

Command	Purpose	Best For
npm run scan	Scan current attention boards.	Current heat
npm run hot -- --limit 20	Rank strongest current topics.	Shortlists
npm run new -- --mode latest	Inspect new/latest/rising/viral topic context.	Fresh attention
npm run spotlight	Load Attention Spotlight and recent signal context.	Spotlight review

Learning And Topic Commands

Command	Purpose	Best For
npm run learning	Load learning summary, topics, outcomes, and resources.	Historical patterns
npm run topics -- --search <query>	Search narrative memory.	Narrative research
npm run topic -- --topic <topic-id>	Inspect one topic and its signal context.	Deep topic review

Token And Artifact Commands

Command	Purpose	Best For
npm run token -- --address <token-address>	Run fast token analysis.	Quick checks
npm run token:deep -- --address <token-address>	Run deeper token due diligence.	Safety and conviction review
npm run thesis -- --topic <topic-id>	Convert a topic into a narrative thesis.	Research notes
npm run risk -- --topic <topic-id>	Produce a risk checklist for a topic.	Invalidation rules
npm run watch -- --topic <topic-id>	Produce a watch plan.	Follow-up monitoring
npm run fetch -- --path /api/<animeme-path>	Fetch an allowed ANIMEME public route.	Advanced read-only inspection

---

Demo Playbooks

Demo 1: First-Run Agent Onboarding

Use the ANIMEME skills. Show me what you can do, then run the default demo flow.

Expected flow:

npm run doctor
npm run demo

The agent should return:

• what ANIMEME can do
• the strongest current Attention Read
• why it matters
• what could invalidate it
• which artifact command to run next

Demo 2: Current Trending Narrative

npm run answer -- --prompt "What narrative is trending right now?"

Expected answer:

• direct answer first
• current Attention Read
• catalyst summary
• crowd-state read
• lead token context when available
• warning or missing-data note
• next prompt

Demo 3: Narrative Explanation

npm run answer -- --prompt "What is the LUNCHMONEY narrative about?"

Expected answer:

• plain-language explanation
• whether the topic is live, stale, or only historical
• ANIMEME learning context
• Spotlight context when available
• what to inspect next

Demo 4: Token Research

npm run answer -- --prompt "Analyze token <token-address>"

Expected answer:

• score
• verdict
• confidence
• live attention status
• learning status
• token context
• strengths
• warnings
• hard stops
• missing data
• next command

Demo 5: Thesis, Risk, And Watch Bundle

npm run scan
npm run thesis -- --topic <topic-id>
npm run risk -- --topic <topic-id>
npm run watch -- --topic <topic-id>

Expected output:

• one narrative thesis
• invalidation rules
• hard-stop checklist
• watch conditions
• generated JSON and Markdown artifacts

---

Output Contract

Good ANIMEME agent answers follow the same shape:

Verdict: watch
Confidence: medium
Score: 54/100

Why it matters:
- The topic has live ANIMEME attention.
- The narrative is easy to retell.
- The learning match is partial.

Warnings:
- Required context is incomplete.
- Crowd-state confirmation is not fully cleared.

Next:
- Run a thesis on the matched topic.
- Keep the token in watch mode until missing context is resolved.

Every good answer should include:

Field	Required?	Why
Direct answer	Yes	The user should not wait for the conclusion.
ANIMEME context used	Yes	Shows what the agent actually checked.
Signal read	Yes	Explains why the topic or token matters.
Warnings	Yes	Prevents hype from replacing judgment.
Missing data	Yes	Missing data is not bullish.
Next command	Yes	Keeps the workflow actionable.

---

Artifact Contract

Commands that generate research output write artifacts under artifacts/.

artifacts/
  2026-05-01T02-17-52-895Z-token-<address>.json
  2026-05-01T02-17-52-895Z-token-<address>.md

Artifact	Purpose
JSON	Machine-readable payload for agents, scripts, and audits.
Markdown	Human-readable summary for review and sharing.

Generated artifacts are advisory and user-controlled. The repository ignores generated artifacts by default, except artifacts/.gitkeep.

---

Token Intelligence Framework

token:deep creates an ANIMEME token-intelligence read.

Inputs:

• token address
• live ANIMEME attention match
• ANIMEME topic context
• ANIMEME learning match
• ANIMEME token context
• concentration and crowding signals
• strength, warning, hard-stop, and missing-data checks

Verdicts:

Verdict	Meaning	Agent Action
researchable	The setup has enough clean context to continue research.	Write thesis, compare with Spotlight, define watch rules.
watch	The setup is mixed or incomplete.	Keep observing and require more proof.
high-risk	Attention is weak or safety context is poor/incomplete.	Avoid escalation unless the user has separate evidence.
avoid	A blocking concentration, manipulation, or integrity risk is present.	Stop escalation and explain the blocking risk.

Never say a token is guaranteed safe.

---

Topic Intelligence Framework

Topic-level work is based on attention, narrative readability, flow, token surface, Spotlight context, and learning memory.

Dimension	Strong Sign	Weak Sign
Attention	Repeated board visibility and strong score	One stale appearance
Catalyst	Clear reason people can repeat	Vague ticker-only noise
Narrative	One-sentence retellability	Confusing context
Token Surface	Visible lead token and coherent basket	Broken or scattered token surface
Spotlight Context	Has signal history	No Spotlight context
Learning Context	Similar prior patterns exist	No historical comparison
Risk Context	Warnings are explicit and bounded	Missing data hidden as confidence

---

Repository Layout

.
+-- .agents/
|   +-- skills/
|       +-- animeme-data/
|       |   +-- SKILL.md
|       +-- animeme-token-intelligence/
|           +-- SKILL.md
+-- artifacts/
|   +-- .gitkeep
+-- docs/
|   +-- data-catalog.md
|   +-- demo-prompt-playbook.md
|   +-- token-intelligence-playbook.md
+-- memory/
|   +-- README.md
+-- src/
|   +-- CLI, clients, routing, and intelligence logic
+-- AGENTS.md
+-- CLAUDE.md
+-- opencode.json
+-- package.json
+-- tsconfig.json

---

Safety Model

Allowed:

• read ANIMEME public intelligence
• analyze and score
• write generated artifacts
• summarize uncertainty and missing data
• produce thesis, risk, and watch plans

Blocked:

• trade or swap
• sign transactions
• request private keys
• request seed phrases
• store credentials, cookies, exported sessions, or wallet material
• mutate production systems
• claim a token is guaranteed safe

All output is research, not financial advice.

---

Documentation Boundary

This public repository documents ANIMEME as the only public intelligence surface.

Public docs must not expose, name, map, or describe non-ANIMEME endpoints. Agents should treat every workflow as ANIMEME intelligence unless a local operator has separately configured private enrichment outside the public documentation surface.

Allowed public language:

• ANIMEME public intelligence
• ANIMEME live attention
• ANIMEME token context
• ANIMEME narrative memory
• ANIMEME spotlight signals
• ANIMEME learning archive
• ANIMEME artifact output

Avoid public documentation that exposes:

• non-ANIMEME endpoint URLs
• non-ANIMEME route paths
• non-ANIMEME credential names
• adapter internals
• private infrastructure topology
• account, wallet, signing, or trading instructions

The public story is simple: agents connect to ANIMEME, ANIMEME returns structured attention intelligence, and the agent turns that intelligence into a clear research answer.

---

Development

npm install
npm run typecheck
npm run doctor
npm run demo
npm run token:deep -- --address <token-address>

When extending the kit:

• keep public docs ANIMEME-only
• add new workflows through the CLI before documenting them
• keep generated files under artifacts/
• document user-facing behavior in README.md, AGENTS.md, and docs/
• do not expose non-ANIMEME endpoint details in public docs
• keep every workflow read-only unless the product explicitly changes scope

---

FAQ

Is this the ANIMEME product?

No. The product is animeme.app. This repository is the public Agent Skill and read-only CLI layer.

Does this trade?

No. It is read-only.

Does this need wallet credentials?

No. Never provide private keys, seed phrases, exported sessions, or wallet material.

Can agents fetch arbitrary websites?

No. Public fetch workflows are constrained to ANIMEME public intelligence.

What happens when data is missing?

The CLI reports missing data explicitly and lowers confidence. Missing data is never treated as bullish.

Is the score financial advice?

No. The score is an agent research heuristic for deciding what to inspect next.

---

Short Version

npx skills add animeme99/Animeme-Agent
git clone https://github.com/animeme99/Animeme-Agent.git
cd Animeme-Agent
npm install
npm run doctor
npm run demo
npm run token:deep -- --address <token-address>

ANIMEME Agent Skill brings public ANIMEME intelligence into your own agent so it can explain attention, catalyst, crowd state, confirmation, and risk before the chart makes the move obvious.