Third Brain V8.1 Skills: LLM wiki kit by Markk-gao

What it is

agent wiki +engineering skills

— the description by Mark393295827

AutoGineer overview, from the brain’s own files

Third Brain V8.1 provides a structured, verification-first operating model for AI agents, treating the LLM as a CPU, context as RAM, and an Obsidian vault as a durable disk. It offers 21 reusable agent skills designed to manage knowledge ingestion, daily loops, research, and multi-agent orchestration. The system enforces strict provenance, governance, and quality gates, ensuring that agent actions are backed by evidence and that knowledge compounds over time through a structured taxonomy.

How it works

Memory model: LLM wiki

AutoGineer overview, from the brain’s own files

The system organizes information into a structured folder hierarchy including sources, concepts, entities, and maps. It uses a 5-stage worker assembly pipeline (Ingest, Cognitive, GraphWeaver, Governance, Deliverable) to process information. Skills are executed as programs within a harness that provides scheduling, permissions, and observability. It relies on a transactional runtime to manage state and ensure that all knowledge updates are verified and linked within the Obsidian vault.

Folders
Folder	Role	Files
.github	agent-config	5
assets	attachments	5
commands	agent-instructions	4
docs	docs	21
examples	examples	10
hooks	hooks	1
skills	agent-instructions	40
sources	raw-sources	13
tests	tests	8
tools	scripts	51
adapters	folder	2
artifacts	folder	28
contracts	folder	18
core	folder	6
experiments	folder	7
maps	folder	1
outreach	folder	46
system	folder	33
workflows	folder	1

Requirements

Works with: Claude Code, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Amp, Cline, JetBrains Junie, Devin (ex-Windsurf)

Set it up

This brain runs code automatically: 2 scripts. Review install.ps1, install.sh before opening the folder in Claude Code. No step below turns a hook on for you.
- install.ps1: pwsh install.ps1
- install.sh: sh install.sh

Clone the repository at the commit AutoGineer scanned:

git clone https://github.com/Mark393295827/third-brain-v7-skills.git && cd third-brain-v7-skills && git checkout 5a645148a108d71c76521715117b737ae60b1423

Claude Code reads CLAUDE.md from this folder automatically.
cd third-brain-v7-skills && claude
AutoGineer scanned these files at commit 5a64514.
First committed by Mark393295827. Published in Mark393295827/third-brain-v7-skills. Licence: MIT License.

Usage

AutoGineer overview, from the brain’s own files

Ingesting research papers or articles into a linked wiki
Running daily knowledge loops to compound insights
Enforcing verification gates before claiming task completion
Orchestrating multi-agent teams with isolated write scopes
Managing context budgets and state checkpoints for long-running tasks

“Use wiki-ingest on this source. Stage an immutable source, author a governed concept candidate, plan navigation updates, and return a verification summary.”: The agent creates an immutable note in the sources folder, updates the relevant concept pages in the wiki, and provides a summary of the verification steps taken.
“Run the daily-okr skill for today.”: The agent executes the 7-KR cycle, generating a daily insight, updating the wiki, and producing a receipt for the completed work.

Documentation by Mark393295827

Third Brain V8.1 Skills

Verification-first Agent Skills for Codex OS, with optional compatibility adapters for Claude Code, Gemini, Cursor, and Windsurf. Build a persistent knowledge operating system with Obsidian provenance, scheduled loops, bounded static dependency graphs, context management, and multi-agent orchestration.

Install 21 reusable agent skills for ingesting sources, auditing workflows, compiling interlinked wikis, running daily knowledge loops, verifying claims before shipping, managing token costs, engineering bounded loops and static dependency graphs, and orchestrating multi-agent teams.

Use this if:

Your AI agent keeps forgetting context between sessions
You want to enforce verification gates before "done" claims
You need to structure knowledge as linked pages (like Obsidian) instead of scattered chat history
You're building agent workflows that should learn and compound over time
You want Obsidian wiki knowledge to update skills, SOPs, schemas, and automation through supervised promotion gates

Image: GitHub stars Image: License: MIT Image: Codex OS Image: Claude Code Image: Cursor Image: PRs Welcome

Quick Start

1. Clone & Install

git clone https://github.com/Mark393295827/third-brain-v5-skills.git
cd third-brain-v5-skills
bash install.sh codex  # Codex is the primary host/kernel

Windows PowerShell uses .\install.ps1 codex. Compatibility adapters require an explicit non-Codex target.

2. Try a Skill

Paste this into Codex:

Use wiki-ingest on this source. Stage an immutable source, author a governed concept candidate, plan navigation updates, and return a verification summary. Treat entity authoring as a separate reviewed capability.

Then paste a URL, article, PDF text, or any source you want to capture.

3. Next Steps

See the 3-Minute Quickstart for a complete walkthrough.

Full guide: GUIDE.md

Command centre and distribution: V8.1 Codex Agentic OS command centre. Start it with python tools/agentic_os_server.py --state-root .agentic-os-state; the host is localhost-only by default and requires T02 runtime receipts. Claude/Gemini/Cursor/Windsurf remain compatibility adapters, not the primary runtime.

中文历史进阶手册：V7 最大潜力使用手册（参考）

🔥 V8.1 多智能体 Worker 流水线手册 & Worker Flows 规范：事务式流水线（Stage → Author → Graph Plan → Govern → Commit/Postcheck → optional Deliverable），归档只发生在验证提交之后。

The Problem

Scenario	Before	After
Research PDF	Summarized once, then forgotten	`wiki-ingest`: source notes, concept pages, linked wiki
Coding session	Agent: "I fixed it" (no proof)	`verify-before-claim`: requires test output + exit code before any claim
Daily work	Tasks and ideas scatter across chat	`daily-okr`: one insight, one wiki update, one action, one output, daily score

V8.1 Operating Model

V8.1 keeps the wiki as durable disk and governance, gives every skill a profile-aware execution contract, and makes python -m tools.worker_flow.cli the only canonical transactional runtime:

Input -> Source -> Wiki compile -> Daily loop -> Agent/Wiki flywheel -> Skill/SOP upgrade -> Verification

The upgrade adds seven hard defaults:

Profiles match risk: one-shot, stateful, loop, and high-risk work load only the controls they need.
Source provenance stays immutable: raw source notes and block refs remain the audit layer.
Loops have contracts: Trigger -> Execute -> Verify -> State, with budgets and recovery.
Context is zero-overhead by default: hot paths load only what the task needs.
Automation is bounded: scans and queues can run unattended; semantic writes stay supervised.
Teams need ownership: multi-agent work requires write scopes, IPC, join gates, cleanup, and evidence.
Rules promote through evidence: wiki insights become schema or skill rules only after repeated support and a cheap check.

See the V8.1 Worker Flow, the Agent Skills standard, and the historical V7 release notes.

V8.1 keeps graph-engineering bounded to static DAGs; dynamic expansion and cyclic graphs remain out of scope.

Core Skills (Start Here)

🧠 Knowledge & Verification

Skill	What it does
wiki-ingest	Preserve an immutable source, compile linked understanding, enforce promotion boundaries, and verify every touched file.
verify-before-claim	Iron rule: no completion claim without fresh, scope-matched evidence; consequential actions also require independent verification, approval, and rollback.

📅 Daily Loop

Skill	What it does
daily-okr	Execute a 7-KR cycle: Input → Cognition → Wiki → Behavior → Creativity → Output → Feedback, with an artifact or receipt for every completed KR.

🎯 Behavior & Creativity

Skill	What it does
behavior-design	Turn outcomes into observable minimum behavior, cues, SOPs, recovery, evidence, and review.
creativity-engine	Generate mechanism-diverse options, rank evidence gaps, and define three bounded minimum experiments.

🔬 Research & Quality

Skill	What it does
deep-research	STOW-compatible research harness with evidence trails, source/claim ledgers, privacy checks, and direct handoff to wiki-ingest.
session-learn	Extract knowledge patterns from sessions — concepts, entities, corrections, patterns, ideas, decisions, gaps. Closure protocol ensures learning feeds back into the wiki.

📊 Context & Engineering

Skill	What it does
context-manager	Derive budgets at runtime, checkpoint state, compact with KEEP/SUMMARIZE/DROP/RETRIEVE, and route by capability.
loop-engineering	Control temporal depth: turn repeatable work into a bounded Trigger -> Execute -> Verify -> State loop with durable state, finite budgets, and explicit stop/recovery rules.
graph-engineering	Control dependency width: validate bounded static DAGs with explicit dependencies, independent branches, typed joins, and node-local recovery.
agentic-engineering	Design agent workflows as spec-driven macro actions with quality ceilings, verification gates, delegated-action boundaries, and state checkpoints.
harness-engineering	Provide the runtime scheduler, permissions, leases, tools, provenance, and observability needed to execute agent workflows safely.
agent-teams-command	Admit and command multi-agent processes with exclusive ownership, typed IPC, isolated writes, serial integration, evidence, and cleanup.

Complete Skill List

📥 Ingestion & Knowledge

wiki-ingest — Stage immutable sources, author governed concept candidates, plan graph updates, and hand off optional entity authoring for separate review
knowledge-ops — Multi-layer knowledge management: classify, deduplicate, preserve evidence hierarchy, vector + Markdown retrieval
wiki-lint — Health check: P0/P1 graph health, frontmatter, source refs, wikilink density, provenance debt

🔄 Daily Workflow

daily-okr — 7-KR evidence cycle with minimum-day degradation and durable closeout
cognitive-compile — 8-section framework: Question → Facts → Concepts → Pattern Recognition → Conflict Detection → Hypothesis Generation

🎨 Behavior & Creativity

behavior-design — Outcome → minimum behavior → cue → SOP → evidence → review
creativity-engine — Mechanism-diverse combinations, scoring, and falsifiable minimum tests

🔬 Research & Quality

deep-research — STOW-compatible research harness, evidence trails, source/claim ledgers
verify-before-claim — Fresh scope-matched evidence, independent checks, approval, rollback

🔄 Learning & Flow

session-learn — 7 signal types: concepts, entities, corrections, patterns, ideas, decisions, gaps
project-flow-ops — Execution flow: triage, plan, track, review across projects

📊 Context

context-manager — Runtime budgets, checkpoint replay, compaction, retrieval, capability routing

Utility command: commands/token-cost-tracker.md estimates, logs, and reports token usage; it is not one of the 21 Agent Skills.

🏗️ Engineering

loop-engineering — Temporal depth through bounded loop contracts, independent verification, finite budgets, and stop/recovery rules
graph-engineering — Dependency width through bounded static DAGs, explicit branches, typed joins, and node-local recovery
agentic-engineering — Spec-driven macro actions, quality ceilings, verification gates, state checks
harness-engineering — Runtime scheduler, permissions, system-call tools, delegated gates, provenance, observability
agent-teams-command — Multi-agent process ownership, IPC, integration, async budgets, and evidence gates

💼 Strategy & Operations

startup-evaluation — Startup health: entrepreneurship, VC 5T, PMF, runway, team, unit economics
anthropic-os — Governed operating-system experiments: Four-C, 3B, prediction errors, permission ladder
ai-six-sigma-property-os — AI + Ontology + DMAIC for property work orders, dispatch, quotes, evidence

How These Skills Fit Together

                        LLM (CPU) + Context (RAM) + Wiki (Disk)

External Sources ──→ wiki-ingest + knowledge-ops ──→ Knowledge Layers
                            ↓
                    Daily Loop (daily-okr)
                    /    /    \    \    \
               Input  Cognition  Wiki  Behavior  Creativity
                            |
                        Output → Feedback
                            ↓
                    session-learn (extract patterns)
                            ↓
                    verify-before-claim (quality gate)
                             ↓
             Loop depth / Graph dependency width
                             ↓
                Multi-agent teams (agent-teams-command)

The system is a closed loop: ingest sources → process daily → extract learning → verify claims → promote rules → scale to teams.

Architecture & Design

Third Brain V8.1 treats agents as LLM OS processes:

LLM = CPU
Context = RAM
Wiki/Obsidian = Disk
Tools = System calls
Skills = Executable programs
Harness = Kernel
Agent teams = Processes

Design Layers

Layer	Principle	Skills
🧠 Knowledge OS	Capture, structure, lint, and promote knowledge over time	wiki-ingest, knowledge-ops, wiki-lint
⚡ Daily Loop	Close the knowledge-to-action cycle every day	daily-okr, cognitive-compile
🎯 Behavior & Creativity	Turn knowledge into habits and novel ideas	behavior-design, creativity-engine
🔬 Research & Quality	Verify before claiming, research with rigor	deep-research, verify-before-claim
🔄 Continuous Learning	Extract patterns from every session	session-learn, project-flow-ops
📊 Context	Manage hot context, durable checkpoints, and runtime budgets	context-manager
🏗️ Engineering	Design bounded loops, static dependency graphs, harnesses, agent workflows, and multi-agent systems	loop-engineering, graph-engineering, agentic-engineering, harness-engineering, agent-teams-command
💼 Strategy & Operations	Evaluate startups, design AI quality systems	startup-evaluation, anthropic-os, ai-six-sigma-property-os

Engineering boundaries are deliberate: Loop Engineering controls temporal depth; Graph Engineering controls dependency width; Agent Teams Command controls process ownership, IPC, and integration; Harness Engineering supplies the runtime scheduler, permissions, and observability. A graph node may contain a Loop or Agent Team, but V8.1 does not support dynamic or cyclic graphs.

Skill Adoption Path

Start small. Add skills as you need them.

Stage	Core Skills	Unlock When
Week 1	`wiki-ingest` + `verify-before-claim`	You can ingest 1 source/day + every claim has fresh evidence
Weeks 2-4	+ `daily-okr` + `session-learn`	Daily artifacts and verified learnings feed back to the wiki
Month 2+	+ `cognitive-compile` + `behavior-design` + `creativity-engine`	Wiki has 50+ pages or repeated decisions need synthesis
Month 3+	+ `knowledge-ops` + `loop-engineering` + `harness-engineering` + `agentic-engineering`	Retrieval, looping reliability, permissions, or delegated actions become bottlenecks
Dependency graph	+ `graph-engineering` after `loop-engineering`	Explicit dependencies, independent branches, typed joins, or node-local recovery create more value than orchestration and review cost
Multi-agent	+ `agent-teams-command` + `project-flow-ops`	Work splits into separate owners with clear integration gates
Strategy	+ `startup-evaluation` + `anthropic-os` + `deep-research`	Need startup health, market, operating-system decisions
Operations	+ `ai-six-sigma-property-os`	Need measurable service quality, dispatch, evidence loops

Installation

For Codex CLI (Primary host/kernel)

# Personal skills (available across all projects)
bash install.sh codex

For Cursor

bash install.sh cursor

For Windsurf

bash install.sh windsurf

Compatibility adapters: Claude Code / Gemini CLI

bash install.sh claude
bash install.sh gemini

Full guide: GUIDE.md

Example Workflows

Each workflow is copyable into your agent. See examples/ for complete, verified examples:

3-minute quickstart — Fastest path to useful output
Research PDF to wiki — Turn a source into linked pages
Verified code session — Use verification gates before claiming a fix
Daily knowledge loop — Run a compact daily OKR workflow
Startup evaluation sprint — Turn an idea into validated assumptions

Wiki Structure

Skills default to STOW paths (configurable via system/config.md):

sources/          ← Immutable source notes (articles, PDFs, transcripts)
wiki/
├── concepts/     ← 13 contracted domains
├── entities/     ← people, companies, funds-investors, products, orgs
└── outputs/      ← gmail-digests, evaluations, compilations
maps/
├── domain-mocs/
├── system-indexes/
├── project-maps/
└── canvases/
system/           ← Deployed contracts, config, schema, templates, runs

Tools & References

Resource	Purpose
tools/index.html	V8.1 visual skill navigator
tools/token-calculator.html	Historical V5 cost-model calculator; not a runtime authority
GUIDE.md	Full installation & troubleshooting
V7 最大潜力使用手册（历史参考）	Earlier profile、知识飞轮、Agent 工程、治理和 30 天采用路径
CLAUDE.md	Claude Code setup
CONTRIBUTING.md	How to contribute skills

What This Is NOT

Not a chat wrapper. Skills are executable prompts that agents follow; they're not productivity theater.
Not a productivity tool with 100 metrics. Daily OKR has seven causally linked KRs with evidence.
Not an all-in-one framework. Pick skills incrementally. You don't need all 20 to start.
Not prescriptive. Adapt paths, frontmatter, and skill triggers to your workflow.

Philosophy

Three core principles:

Verification first: No "done" without proof. No claims without evidence. Expected value over confidence.
Knowledge compounds: Every session should improve the wiki, not scatter across chat history.
Closed loops: Ingest → Process → Learn → Scale → Verify. No loose ends.

Agent Skills Format — Open specification for agent skills
llm-wiki-agent — Original STOW pattern implementation by SamurAIGPT
Karpathy LLM OS — Conceptual framework

Contributing

Bug reports, skill improvements, and PRs are welcome. See CONTRIBUTING.md and CHANGELOG.md.

License

MIT — see LICENSE.

Transparency Note

This project includes growth and outreach tools (in tools/ and outreach/) designed to help the repository reach users via GitHub search and Awesome lists. These tools are optional, dry-run-by-default, and separate from the core skills. If you're installing skills for your own workflow, you don't need them.

The skill frameworks and philosophy are genuine. Use what works for you; ignore what doesn't.

tools/token-calculator.html

Shown as plain text.

<!DOCTYPE html>
<html lang="zh">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>V5.0 Token成本计算器</title>
<style>
  @import url('https://fonts.googleapis.com/css2?family=JetBrains+Mono:wght@300;400;600;700&family=Noto+Sans+SC:wght@300;400;500;700&display=swap');

  :root {
    --bg: #080c14;
    --surface: #0f1623;
    --surface2: #161f30;
    --surface3: #1c2840;
    --border: #1e3a5f;
    --cyan: #00d4ff;
    --green: #00ff94;
    --amber: #ffb800;
    --red: #ff4757;
    --purple: #a855f7;
    --pink: #ff6eb4;
    --orange: #ff8c42;
    --text: #e2e8f0;
    --text-dim: #94a3b8;
    --text-muted: #4a5568;
  }

  * { box-sizing: border-box; margin: 0; padding: 0; }

  body {
    background: var(--bg);
    font-family: 'Noto Sans SC', sans-serif;
    color: var(--text);
    min-height: 100vh;
    padding: 20px;
  }

  /* HEADER */
  .header {
    text-align: center;
    margin-bottom: 24px;
    padding-bottom: 16px;
    border-bottom: 1px solid rgba(0,212,255,0.2);
  }
  .header h1 {
    font-size: 20px; font-weight: 700; letter-spacing: 2px;
    background: linear-gradient(135deg, var(--cyan), var(--green));
    -webkit-background-clip: text; -webkit-text-fill-color: transparent;
  }
  .header p {
    font-size: 10px; color: var(--text-muted);
    font-family: 'JetBrains Mono'; letter-spacing: 3px; margin-top: 4px;
  }

  /* LAYOUT */
  .layout {
    display: grid;
    grid-template-columns: 340px 1fr;
    gap: 16px;
    align-items: start;
  }

  /* PANEL */
  .panel {
    background: var(--surface);
    border: 1px solid var(--border);
    border-radius: 8px;
    overflow: hidden;
  }
  .panel-header {
    padding: 8px 14px;
    font-size: 10px; font-weight: 700; letter-spacing: 2px;
    font-family: 'JetBrains Mono';
    border-bottom: 1px solid var(--border);
  }
  .panel-body { padding: 14px; }

  /* SECTION LABELS */
  .section-label {
    font-size: 9px; font-weight: 700; letter-spacing: 2px;
    font-family: 'JetBrains Mono'; color: var(--text-muted);
    margin: 12px 0 8px; padding-bottom: 4px;
    border-bottom: 1px solid var(--border);
  }
  .section-label:first-child { margin-top: 0; }

  /* SLIDER ROW */
  .slider-row {
    margin-bottom: 10px;
  }
  .slider-top {
    display: flex; justify-content: space-between; align-items: center;
    margin-bottom: 4px;
  }
  .slider-name { font-size: 10px; color: var(--text-dim); }
  .slider-val {
    font-size: 11px; font-weight: 700; font-family: 'JetBrains Mono';
    color: var(--cyan); min-width: 60px; text-align: right;
  }
  input[type=range] {
    width: 100%; height: 4px;
    -webkit-appearance: none; background: var(--surface3);
    border-radius: 2px; outline: none; cursor: pointer;
  }
  input[type=range]::-webkit-slider-thumb {
    -webkit-appearance: none;
    width: 14px; height: 14px; border-radius: 50%;
    background: var(--cyan); cursor: pointer;
    box-shadow: 0 0 8px rgba(0,212,255,0.5);
  }
  input[type=range].amber::-webkit-slider-thumb { background: var(--amber); box-shadow: 0 0 8px rgba(255,184,0,0.5); }
  input[type=range].green::-webkit-slider-thumb { background: var(--green); box-shadow: 0 0 8px rgba(0,255,148,0.5); }
  input[type=range].purple::-webkit-slider-thumb { background: var(--purple); box-shadow: 0 0 8px rgba(168,85,247,0.5); }

  /* MODEL SELECT */
  .model-grid {
    display: grid; grid-template-columns: 1fr 1fr;
    gap: 6px; margin-top: 4px;
  }
  .model-btn {
    background: var(--surface2); border: 1px solid var(--border);
    border-radius: 5px; padding: 7px 8px;
    cursor: pointer; text-align: center;
    transition: all 0.15s;
  }
  .model-btn:hover { border-color: var(--cyan); }
  .model-btn.active { border-color: var(--cyan); background: rgba(0,212,255,0.08); }
  .model-btn .m-name { font-size: 9px; font-weight: 700; font-family: 'JetBrains Mono'; color: var(--text); }
  .model-btn .m-price { font-size: 8px; color: var(--text-muted); margin-top: 2px; }
  .model-btn.active .m-name { color: var(--cyan); }

  /* COST DISPLAY AREA */
  .cost-col { display: flex; flex-direction: column; gap: 12px; }

  /* STAGE BREAKDOWN */
  .stage-card {
    background: var(--surface);
    border: 1px solid var(--border);
    border-radius: 8px; overflow: hidden;
  }
  .stage-header {
    display: flex; align-items: center; justify-content: space-between;
    padding: 8px 14px;
    border-bottom: 1px solid var(--border);
  }
  .stage-title {
    display: flex; align-items: center; gap: 8px;
    font-size: 10px; font-weight: 700; font-family: 'JetBrains Mono';
  }
  .stage-num {
    width: 20px; height: 20px; border-radius: 4px;
    display: flex; align-items: center; justify-content: center;
    font-size: 9px; font-weight: 700; flex-shrink: 0;
  }
  .stage-cost-badge {
    font-size: 11px; font-weight: 700; font-family: 'JetBrains Mono';
    padding: 3px 10px; border-radius: 4px;
  }
  .stage-body { padding: 10px 14px; }

  /* TOKEN ROW */
  .token-row {
    display: grid; grid-template-columns: 1fr 80px 80px 80px;
    gap: 4px; align-items: center;
    padding: 5px 0; border-bottom: 1px solid rgba(30,58,95,0.3);
    font-size: 9px;
  }
  .token-row:last-child { border-bottom: none; }
  .token-row .t-name { color: var(--text-dim); }
  .token-row .t-val { font-family: 'JetBrains Mono'; text-align: right; color: var(--text-dim); font-size: 8px; }
  .token-row .t-cost { font-family: 'JetBrains Mono'; text-align: right; font-weight: 600; font-size: 9px; }
  .token-row-header {
    display: grid; grid-template-columns: 1fr 80px 80px 80px;
    gap: 4px; padding-bottom: 5px; margin-bottom: 3px;
    border-bottom: 1px solid var(--border);
    font-size: 8px; color: var(--text-muted);
    font-family: 'JetBrains Mono'; letter-spacing: 1px;
  }
  .token-row-header span { text-align: right; }
  .token-row-header span:first-child { text-align: left; }

  /* PROGRESS BAR IN STAGE */
  .token-bar {
    margin-top: 8px; background: var(--surface2);
    border-radius: 3px; height: 6px; overflow: hidden;
    display: flex;
  }
  .token-bar-seg { height: 100%; transition: width 0.3s; }

  /* SUMMARY TOTALS */
  .summary-card {
    background: linear-gradient(135deg, rgba(0,212,255,0.04), rgba(0,255,148,0.04));
    border: 1px solid rgba(0,212,255,0.3);
    border-radius: 8px; padding: 14px;
  }
  .summary-title {
    font-size: 10px; font-weight: 700; letter-spacing: 2px;
    font-family: 'JetBrains Mono'; color: var(--cyan); margin-bottom: 12px;
  }
  .summary-grid {
    display: grid; grid-template-columns: repeat(4, 1fr); gap: 8px;
    margin-bottom: 12px;
  }
  .summary-stat {
    background: var(--surface);
    border: 1px solid var(--border);
    border-radius: 6px; padding: 10px;
    text-align: center;
  }
  .summary-stat-num {
    font-size: 16px; font-weight: 700; font-family: 'JetBrains Mono';
    margin-bottom: 3px;
  }
  .summary-stat-label {
    font-size: 8px; color: var(--text-muted); font-family: 'JetBrains Mono'; letter-spacing: 1px;
  }
  .optimization-tips {
    background: var(--surface2); border-radius: 5px; padding: 10px;
  }
  .opt-title { font-size: 9px; font-weight: 700; color: var(--green); font-family: 'JetBrains Mono'; margin-bottom: 6px; }
  .opt-item { font-size: 9px; color: var(--text-dim); padding: 2px 0; display: flex; gap: 6px; }
  .opt-item::before { content: '→'; color: var(--green); flex-shrink: 0; }
  .opt-saving { font-family: 'JetBrains Mono'; color: var(--green); font-weight: 600; }

  /* FUNNEL VISUAL IN COST */
  .funnel-cost {
    display: flex; gap: 6px; align-items: center; margin-top: 8px;
    flex-wrap: wrap;
  }
  .funnel-step {
    background: var(--surface2); border: 1px solid var(--border);
    border-radius: 4px; padding: 4px 8px; font-size: 8px;
    font-family: 'JetBrains Mono'; color: var(--text-dim); text-align: center;
  }
  .funnel-step .fs-num { font-size: 10px; font-weight: 700; color: var(--orange); }
  .funnel-arrow { color: var(--text-muted); font-size: 10px; }

  /* COLOR HELPERS */
  .c-cyan { color: var(--cyan); }
  .c-green { color: var(--green); }
  .c-amber { color: var(--amber); }
  .c-red { color: var(--red); }
  .c-purple { color: var(--purple); }
  .c-pink { color: var(--pink); }
  .c-orange { color: var(--orange); }
  .bg-cyan { background: var(--cyan); }
  .bg-green { background: var(--green); }
  .bg-amber { background: var(--amber); }
  .bg-purple { background: var(--purple); }
  .bg-pink { background: var(--pink); }
  .bg-orange { background: var(--orange); }

  .note { font-size: 8px; color: var(--text-muted); font-family: 'JetBrains Mono'; margin-top: 6px; line-height: 1.5; }
</style>
</head>
<body>
<div class="header">
  <h1>V5.0 Pipeline Token成本计算器</h1>
  <p>PER-STAGE TOKEN COST BREAKDOWN · FULL PIPELINE SIMULATION</p>
</div>

<div class="layout">
  <!-- LEFT: CONTROLS -->
  <div>
    <!-- INPUT VOLUME -->
    <div class="panel" style="margin-bottom:12px;">
      <div class="panel-header" style="color:var(--cyan)">① 每日输入量</div>
      <div class="panel-body">
        <div class="section-label">摄取层</div>
        <div class="slider-row">
          <div class="slider-top">
            <span class="slider-name">文章/网页数量</span>
            <span class="slider-val" id="v-articles">20</span>
          </div>
          <input type="range" id="s-articles" min="1" max="100" value="20" oninput="update()">
        </div>
        <div class="slider-row">
          <div class="slider-top">
            <span class="slider-name">平均文章长度 (tokens)</span>
            <span class="slider-val" id="v-article-len">1,500</span>
          </div>
          <input type="range" class="amber" id="s-article-len" min="300" max="5000" step="100" value="1500" oninput="update()">
        </div>
        <div class="slider-row">
          <div class="slider-top">
            <span class="slider-name">视频转录数量</span>
            <span class="slider-val" id="v-videos">3</span>
          </div>
          <input type="range" class="green" id="s-videos" min="0" max="20" value="3" oninput="update()">
        </div>
        <div class="slider-row">
          <div class="slider-top">
            <span class="slider-name">平均转录长度 (tokens)</span>
            <span class="slider-val" id="v-video-len">4,000</span>
          </div>
          <input type="range" class="green" id="s-video-len" min="1000" max="12000" step="500" value="4000" oninput="update()">
        </div>

        <div class="section-label">Layer 2 处理</div>
        <div class="slider-row">
          <div class="slider-top">
            <span class="slider-name">每日主动LLM调用次数</span>
            <span class="slider-val" id="v-calls">10</span>
          </div>
          <input type="range" class="purple" id="s-calls" min="1" max="50" value="10" oninput="update()">
        </div>
        <div class="slider-row">
          <div class="slider-top">
            <span class="slider-name">平均输出长度 (tokens)</span>
            <span class="slider-val" id="v-output-len">800</span>
          </div>
          <input type="range" class="purple" id="s-output-len" min="100" max="4000" step="100" value="800" oninput="update()">
        </div>

        <div class="section-label">Layer 3 行动层</div>
        <div class="slider-row">
          <div class="slider-top">
            <span class="slider-name">深度研究/产出任务数</span>
            <span class="slider-val" id="v-research">2</span>
          </div>
          <input type="range" id="s-research" min="0" max="10" value="2" oninput="update()">
        </div>
        <div class="slider-row">
          <div class="slider-top">
            <span class="slider-name">产出平均长度 (tokens)</span>
            <span class="slider-val" id="v-research-out">3,000</span>
          </div>
          <input type="range" class="amber" id="s-research-out" min="500" max="10000" step="500" value="3000" oninput="update()">
        </div>
      </div>
    </div>

    <!-- MODEL SELECTION -->
    <div class="panel">
      <div class="panel-header" style="color:var(--amber)">② 主模型选择</div>
      <div class="panel-body">
        <div style="font-size:9px;color:var(--text-muted);font-family:'JetBrains Mono';margin-bottom:6px;">Layer 2/3 主要调用模型（per M tokens）</div>
        <div class="model-grid" id="model-grid">
          <div class="model-btn active" onclick="selectModel('haiku', this)">
            <div class="m-name">Claude Haiku</div>
            <div class="m-price">$0.80 / $4.00</div>
          </div>
          <div class="model-btn" onclick="selectModel('sonnet', this)">
            <div class="m-name">Claude Sonnet</div>
            <div class="m-price">$3.00 / $15.00</div>
          </div>
          <div class="model-btn" onclick="selectModel('opus', this)">
            <div class="m-name">Claude Opus</div>
            <div class="m-price">$15.00 / $75.00</div>
          </div>
          <div class="model-btn" onclick="selectModel('gpt4o', this)">
            <div class="m-name">GPT-4o</div>
            <div class="m-price">$2.50 / $10.00</div>
          </div>
          <div class="model-btn" onclick="selectModel('gpt4mini', this)">
            <div class="m-name">GPT-4o mini</div>
            <div class="m-price">$0.15 / $0.60</div>
          </div>
          <div class="model-btn" onclick="selectModel('gemini', this)">
            <div class="m-name">Gemini 1.5 Pro</div>
            <div class="m-price">$1.25 / $5.00</div>
          </div>
        </div>
        <div class="note">* 价格仅供估算参考，请以官方文档为准。漏斗过滤和嵌入使用轻量模型/本地方案。</div>
      </div>
    </div>
  </div>

  <!-- RIGHT: COST BREAKDOWN -->
  <div class="cost-col">

    <!-- STAGE 1: FUNNEL FILTERING -->
    <div class="stage-card">
      <div class="stage-header">
        <div class="stage-title">
          <div class="stage-num bg-orange" style="color:#000">F</div>
          <span class="c-orange">知识漏斗：内容分类过滤</span>
          <span style="font-size:8px;color:var(--text-muted);font-weight:400">(轻量模型 Haiku)</span>
        </div>
        <div class="stage-cost-badge" id="stage-funnel-badge" style="background:rgba(255,140,66,0.15);color:var(--orange)">$0.00</div>
      </div>
      <div class="stage-body">
        <div class="token-row-header">
          <span>步骤</span><span>输入Tokens</span><span>输出Tokens</span><span>成本/天</span>
        </div>
        <div class="token-row">
          <span class="t-name">F1 稀缺性分类 (100%素材)</span>
          <span class="t-val" id="f1-in">-</span>
          <span class="t-val" id="f1-out">-</span>
          <span class="t-cost c-orange" id="f1-cost">-</span>
        </div>
        <div class="token-row">
          <span class="t-name">F2 价值度评分 (60%通过)</span>
          <span class="t-val" id="f2-in">-</span>
          <span class="t-val" id="f2-out">-</span>
          <span class="t-cost c-orange" id="f2-cost">-</span>
        </div>
        <div class="token-row">
          <span class="t-name">F3 时效性裁切 (40%→24%)</span>
          <span class="t-val" id="f3-in">-</span>
          <span class="t-val" id="f3-out">-</span>
          <span class="t-cost c-orange" id="f3-cost">-</span>
        </div>
        <div class="funnel-cost">
          <div class="funnel-step">全量素材<br><span class="fs-num" id="fc-total">0</span><br><span>件</span></div>
          <div class="funnel-arrow">→F1→</div>
          <div class="funnel-step">稀缺信号<br><span class="fs-num" id="fc-f1">0</span><br><span>件 60%</span></div>
          <div class="funnel-arrow">→F2→</div>
          <div class="funnel-step">高价值<br><span class="fs-num" id="fc-f2">0</span><br><span>件 40%</span></div>
          <div class="funnel-arrow">→F3→</div>
          <div class="funnel-step" style="border-color:var(--amber);color:var(--amber)">进行动<br><span class="fs-num" id="fc-f3">0</span><br><span>件 ~12%</span></div>
        </div>
      </div>
    </div>

    <!-- STAGE 2: CONTEXT ASSEMBLY + LLM PROCESSING -->
    <div class="stage-card">
      <div class="stage-header">
        <div class="stage-title">
          <div class="stage-num" style="background:var(--green);color:#000">2</div>
          <span class="c-green">Layer 2：Context组装 + LLM处理</span>
        </div>
        <div class="stage-cost-badge" id="stage-l2-badge" style="background:rgba(0,255,148,0.1);color:var(--green)">$0.00</div>
      </div>
      <div class="stage-body">
        <div class="token-row-header">
          <span>组件</span><span>输入Tokens</span><span>输出Tokens</span><span>成本/天</span>
        </div>
        <div class="token-row">
          <span class="t-name">Schema (P0 · 25% ctx)</span>
          <span class="t-val" id="l2-schema-in">-</span>
          <span class="t-val">—</span>
          <span class="t-cost c-purple" id="l2-schema-cost">-</span>
        </div>
        <div class="token-row">
          <span class="t-name">RAG结果 (P1 · 40% ctx)</span>
          <span class="t-val" id="l2-rag-in">-</span>
          <span class="t-val">—</span>
          <span class="t-cost c-cyan" id="l2-rag-cost">-</span>
        </div>
        <div class="token-row">
          <span class="t-name">Log.md (P2 · 20% ctx)</span>
          <span class="t-val" id="l2-log-in">-</span>
          <span class="t-val">—</span>
          <span class="t-cost c-green" id="l2-log-cost">-</span>
        </div>
        <div class="token-row">
          <span class="t-name">System Prompt (15% ctx)</span>
          <span class="t-val" id="l2-sys-in">-</span>
          <span class="t-val">—</span>
          <span class="t-cost c-amber" id="l2-sys-cost">-</span>
        </div>
        <div class="token-row" style="border-top:1px solid var(--border);margin-top:4px;padding-top:8px;">
          <span class="t-name" style="font-weight:600;color:var(--text)">LLM输出 (×调用次数)</span>
          <span class="t-val">—</span>
          <span class="t-val" id="l2-output-val">-</span>
          <span class="t-cost c-green" id="l2-output-cost">-</span>
        </div>
        <div class="token-bar" id="ctx-bar">
          <div class="token-bar-seg" style="background:var(--purple)" id="bar-schema"></div>
          <div class="token-bar-seg" style="background:var(--cyan)" id="bar-rag"></div>
          <div class="token-bar-seg" style="background:var(--green)" id="bar-log"></div>
          <div class="token-bar-seg" style="background:var(--amber)" id="bar-sys"></div>
        </div>
        <div style="display:flex;gap:10px;margin-top:4px;font-size:8px;font-family:'JetBrains Mono';color:var(--text-muted);">
          <span><span style="color:var(--purple)">■</span> Schema</span>
          <span><span style="color:var(--cyan)">■</span> RAG</span>
          <span><span style="color:var(--green)">■</span> Log</span>
          <span><span style="color:var(--amber)">■</span> System</span>
        </div>
      </div>
    </div>

    <!-- STAGE 3: ACTION LAYER -->
    <div class="stage-card">
      <div class="stage-header">
        <div class="stage-title">
          <div class="stage-num bg-amber" style="color:#000">3</div>
          <span class="c-amber">Layer 3：深度研究 + 产出生成</span>
        </div>
        <div class="stage-cost-badge" id="stage-l3-badge" style="background:rgba(255,184,0,0.1);color:var(--amber)">$0.00</div>
      </div>
      <div class="stage-body">
        <div class="token-row-header">
          <span>步骤</span><span>输入Tokens</span><span>输出Tokens</span><span>成本/天</span>
        </div>
        <div class="token-row">
          <span class="t-name">研究Context (RAG+Schema)</span>
          <span class="t-val" id="l3-ctx-in">-</span>
          <span class="t-val">—</span>
          <span class="t-cost c-amber" id="l3-ctx-cost">-</span>
        </div>
        <div class="token-row">
          <span class="t-name">产出生成 (商业计划/代码/模型)</span>
          <span class="t-val">—</span>
          <span class="t-val" id="l3-out-val">-</span>
          <span class="t-cost c-amber" id="l3-out-cost">-</span>
        </div>
        <div class="token-row">
          <span class="t-name">Closure回流 (摘要压缩写入)</span>
          <span class="t-val" id="l3-closure-in">-</span>
          <span class="t-val" id="l3-closure-out">-</span>
          <span class="t-cost c-amber" id="l3-closure-cost">-</span>
        </div>
      </div>
    </div>

    <!-- STAGE E: EMBEDDING -->
    <div class="stage-card">
      <div class="stage-header">
        <div class="stage-title">
          <div class="stage-num bg-pink" style="color:#000">E</div>
          <span class="c-pink">向量嵌入：Vector DB写入</span>
          <span style="font-size:8px;color:var(--text-muted);font-weight:400">(text-embedding-3-small)</span>
        </div>
        <div class="stage-cost-badge" id="stage-embed-badge" style="background:rgba(255,110,180,0.1);color:var(--pink)">$0.00</div>
      </div>
      <div class="stage-body">
        <div class="token-row-header">
          <span>内容</span><span>Tokens量</span><span>—</span><span>成本/天</span>
        </div>
        <div class="token-row">
          <span class="t-name">漏斗通过素材嵌入 (~12%)</span>
          <span class="t-val" id="emb-content">-</span>
          <span class="t-val">—</span>
          <span class="t-cost c-pink" id="emb-content-cost">-</span>
        </div>
        <div class="token-row">
          <span class="t-name">Closure产出回流嵌入</span>
          <span class="t-val" id="emb-closure">-</span>
          <span class="t-val">—</span>
          <span class="t-cost c-pink" id="emb-closure-cost">-</span>
        </div>
        <div class="note">* 本地方案(ChromaDB+nomic-embed)成本≈$0，但需GPU或接受较慢速度</div>
      </div>
    </div>

    <!-- SUMMARY -->
    <div class="summary-card">
      <div class="summary-title">⊕ 总成本汇总</div>
      <div class="summary-grid">
        <div class="summary-stat">
          <div class="summary-stat-num c-cyan" id="total-daily">$0.00</div>
          <div class="summary-stat-label">每日</div>
        </div>
        <div class="summary-stat">
          <div class="summary-stat-num c-green" id="total-weekly">$0.00</div>
          <div class="summary-stat-label">每周</div>
        </div>
        <div class="summary-stat">
          <div class="summary-stat-num c-amber" id="total-monthly">$0.00</div>
          <div class="summary-stat-label">每月</div>
        </div>
        <div class="summary-stat">
          <div class="summary-stat-num c-red" id="total-yearly">$0.00</div>
          <div class="summary-stat-label">每年</div>
        </div>
      </div>
      <div class="optimization-tips">
        <div class="opt-title">⚡ 优化建议（当前参数）</div>
        <div class="opt-item" id="tip-funnel">漏斗节省: 计算中...</div>
        <div class="opt-item" id="tip-model">模型选择: 计算中...</div>
        <div class="opt-item" id="tip-cache">缓存策略: Schema+System Prompt启用缓存可节省约 <span class="opt-saving" id="tip-cache-val">$0</span>/月</div>
        <div class="opt-item" id="tip-embed">嵌入优化: 本地部署可节省 <span class="opt-saving" id="tip-embed-val">$0</span>/月</div>
      </div>
    </div>

  </div>
</div>

<script>
  // PRICING TABLE (per 1M tokens: [input, output])
  const MODELS = {
    haiku:   { name: 'Claude Haiku',    input: 0.80,  output: 4.00  },
    sonnet:  { name: 'Claude Sonnet',   input: 3.00,  output: 15.00 },
    opus:    { name: 'Claude Opus',     input: 15.00, output: 75.00 },
    gpt4o:   { name: 'GPT-4o',          input: 2.50,  output: 10.00 },
    gpt4mini:{ name: 'GPT-4o mini',     input: 0.15,  output: 0.60  },
    gemini:  { name: 'Gemini 1.5 Pro',  input: 1.25,  output: 5.00  },
  };
  const EMBED_PRICE = 0.02; // per 1M tokens (text-embedding-3-small)
  const HAIKU = MODELS.haiku; // Always use Haiku for filtering

  let selectedModel = 'haiku';

  function selectModel(key, el) {
    selectedModel = key;
    document.querySelectorAll('.model-btn').forEach(b => b.classList.remove('active'));
    el.classList.add('active');
    update();
  }

  function fmt(n) {
    if (n === 0) return '$0.000';
    if (n < 0.001) return '$<0.001';
    if (n < 1) return '$' + n.toFixed(4);
    if (n < 100) return '$' + n.toFixed(3);
    return '$' + n.toFixed(2);
  }

  function fmtTok(n) {
    if (n >= 1000000) return (n/1000000).toFixed(1) + 'M';
    if (n >= 1000) return (n/1000).toFixed(1) + 'K';
    return n.toString();
  }

  function cost(inputTok, outputTok, model) {
    const m = MODELS[model];
    return (inputTok / 1e6) * m.input + (outputTok / 1e6) * m.output;
  }

  function update() {
    // READ INPUTS
    const articles = +document.getElementById('s-articles').value;
    const artLen   = +document.getElementById('s-article-len').value;
    const videos   = +document.getElementById('s-videos').value;
    const vidLen   = +document.getElementById('s-video-len').value;
    const calls    = +document.getElementById('s-calls').value;
    const outLen   = +document.getElementById('s-output-len').value;
    const research = +document.getElementById('s-research').value;
    const resOut   = +document.getElementById('s-research-out').value;

    // UPDATE DISPLAY VALUES
    document.getElementById('v-articles').textContent    = articles;
    document.getElementById('v-article-len').textContent = artLen.toLocaleString();
    document.getElementById('v-videos').textContent      = videos;
    document.getElementById('v-video-len').textContent   = vidLen.toLocaleString();
    document.getElementById('v-calls').textContent       = calls;
    document.getElementById('v-output-len').textContent  = outLen.toLocaleString();
    document.getElementById('v-research').textContent    = research;
    document.getElementById('v-research-out').textContent = resOut.toLocaleString();

    const totalItems = articles + videos;
    const totalRawTokens = articles * artLen + videos * vidLen;

    // FUNNEL PASS-THROUGH
    const f1Pass = Math.round(totalItems * 0.60);
    const f2Pass = Math.round(f1Pass * 0.40 / 0.60); // of remaining after F1 applying 40% overall = ~24%
    const f3Pass = Math.round(totalItems * 0.12);

    // Actually let's do it linearly:
    // F1: 60% of all items pass
    // F2: of those, ~67% pass (so overall 40%)  
    // F3: of those, ~30% pass (so overall ~12%)
    const after_f1 = Math.round(totalItems * 0.60);
    const after_f2 = Math.round(totalItems * 0.40);
    const after_f3 = Math.round(totalItems * 0.12);

    document.getElementById('fc-total').textContent = totalItems;
    document.getElementById('fc-f1').textContent    = after_f1;
    document.getElementById('fc-f2').textContent    = after_f2;
    document.getElementById('fc-f3').textContent    = after_f3;

    // FUNNEL COST (all use Haiku for classification)
    // Each item: ~200 token system prompt + summary of content (~300 tokens) → 20 token output
    const funnelPromptPer = 250; // classifier prompt
    const funnelInputPer  = 300; // content summary
    const funnelOutputPer = 20;  // yes/no + score

    const f1In  = totalItems * (funnelPromptPer + funnelInputPer);
    const f1Out = totalItems * funnelOutputPer;
    const f2In  = after_f1  * (funnelPromptPer + funnelInputPer);
    const f2Out = after_f1  * funnelOutputPer;
    const f3In  = after_f2  * (funnelPromptPer + 100); // shorter check
    const f3Out = after_f2  * funnelOutputPer;

    const f1c = cost(f1In, f1Out, 'haiku');
    const f2c = cost(f2In, f2Out, 'haiku');
    const f3c = cost(f3In, f3Out, 'haiku');
    const funnelTotal = f1c + f2c + f3c;

    document.getElementById('f1-in').textContent   = fmtTok(f1In);
    document.getElementById('f1-out').textContent  = fmtTok(f1Out);
    document.getElementById('f1-cost').textContent = fmt(f1c);
    document.getElementById('f2-in').textContent   = fmtTok(f2In);
    document.getElementById('f2-out').textContent  = fmtTok(f2Out);
    document.getElementById('f2-cost').textContent = fmt(f2c);
    document.getElementById('f3-in').textContent   = fmtTok(f3In);
    document.getElementById('f3-out').textContent  = fmtTok(f3Out);
    document.getElementById('f3-cost').textContent = fmt(f3c);
    document.getElementById('stage-funnel-badge').textContent = fmt(funnelTotal);

    // LAYER 2: CONTEXT ASSEMBLY
    // Context Window: Schema 25%, RAG 40%, Log 20%, System 15% of total ctx window
    // Assume 8000 token context window per call
    const ctxWindow = 8000;
    const schemaTokens = ctxWindow * 0.25;
    const ragTokens    = ctxWindow * 0.40;
    const logTokens    = ctxWindow * 0.20;
    const sysTokens    = ctxWindow * 0.15;

    const l2SchemaTok = schemaTokens * calls;
    const l2RagTok    = ragTokens    * calls;
    const l2LogTok    = logTokens    * calls;
    const l2SysTok    = sysTokens    * calls;
    const l2TotalIn   = ctxWindow    * calls;
    const l2TotalOut  = outLen       * calls;

    const l2SchemaC = cost(l2SchemaTok, 0, selectedModel);
    const l2RagC    = cost(l2RagTok,    0, selectedModel);
    const l2LogC    = cost(l2LogTok,    0, selectedModel);
    const l2SysC    = cost(l2SysTok,    0, selectedModel);
    const l2OutC    = cost(0, l2TotalOut, selectedModel);
    const l2Total   = cost(l2TotalIn, l2TotalOut, selectedModel);

    document.getElementById('l2-schema-in').textContent   = fmtTok(l2SchemaTok);
    document.getElementById('l2-schema-cost').textContent = fmt(l2SchemaC);
    document.getElementById('l2-rag-in').textContent      = fmtTok(l2RagTok);
    document.getElementById('l2-rag-cost').textContent    = fmt(l2RagC);
    document.getElementById('l2-log-in').textContent      = fmtTok(l2LogTok);
    document.getElementById('l2-log-cost').textContent    = fmt(l2LogC);
    document.getElementById('l2-sys-in').textContent      = fmtTok(l2SysTok);
    document.getElementById('l2-sys-cost').textContent    = fmt(l2SysC);
    document.getElementById('l2-output-val').textContent  = fmtTok(l2TotalOut);
    document.getElementById('l2-output-cost').textContent = fmt(l2OutC);
    document.getElementById('stage-l2-badge').textContent = fmt(l2Total);

    // Context bar widths
    document.getElementById('bar-schema').style.width = '25%';
    document.getElementById('bar-rag').style.width    = '40%';
    document.getElementById('bar-log').style.width    = '20%';
    document.getElementById('bar-sys').style.width    = '15%';

    // LAYER 3: RESEARCH + OUTPUT
    const l3CtxIn   = 12000 * research; // larger context for research
    const l3OutTok  = resOut * research;
    const closureIn = Math.round(resOut * 0.3) * research; // closure summary ~30% of output
    const closureOut= Math.round(resOut * 0.1) * research;

    const l3CtxC    = cost(l3CtxIn, 0, selectedModel);
    const l3OutC    = cost(0, l3OutTok, selectedModel);
    const l3ClosureC= cost(closureIn, closureOut, selectedModel);
    const l3Total   = l3CtxC + l3OutC + l3ClosureC;

    document.getElementById('l3-ctx-in').textContent      = fmtTok(l3CtxIn);
    document.getElementById('l3-ctx-cost').textContent    = fmt(l3CtxC);
    document.getElementById('l3-out-val').textContent     = fmtTok(l3OutTok);
    document.getElementById('l3-out-cost').textContent    = fmt(l3OutC);
    document.getElementById('l3-closure-in').textContent  = fmtTok(closureIn);
    document.getElementById('l3-closure-out').textContent = fmtTok(closureOut);
    document.getElementById('l3-closure-cost').textContent= fmt(l3ClosureC);
    document.getElementById('stage-l3-badge').textContent = fmt(l3Total);

    // EMBEDDINGS
    const embContentTok  = after_f3 * 600; // avg content chunk for passed items
    const embClosureTok  = resOut * research;
    const embContentCost = (embContentTok / 1e6) * EMBED_PRICE;
    const embClosureCost = (embClosureTok / 1e6) * EMBED_PRICE;
    const embedTotal     = embContentCost + embClosureCost;

    document.getElementById('emb-content').textContent      = fmtTok(embContentTok);
    document.getElementById('emb-content-cost').textContent = fmt(embContentCost);
    document.getElementById('emb-closure').textContent      = fmtTok(embClosureTok);
    document.getElementById('emb-closure-cost').textContent = fmt(embClosureCost);
    document.getElementById('stage-embed-badge').textContent= fmt(embedTotal);

    // GRAND TOTAL
    const daily  = funnelTotal + l2Total + l3Total + embedTotal;
    const weekly = daily * 7;
    const monthly= daily * 30;
    const yearly = daily * 365;

    document.getElementById('total-daily').textContent  = fmt(daily);
    document.getElementById('total-weekly').textContent = fmt(weekly);
    document.getElementById('total-monthly').textContent= fmt(monthly);
    document.getElementById('total-yearly').textContent = fmt(yearly);

    // OPTIMIZATION TIPS
    const withoutFunnel = cost(totalRawTokens, 0, selectedModel) * calls / after_f3; // rough estimate
    const funnelSaving  = (l2Total / 0.12 * 0.88); // what would cost without funnel
    document.getElementById('tip-funnel').textContent =
      `漏斗节省: ~${Math.round(88)}%素材被过滤，估计节省 ${fmt(funnelSaving)}/天 的无效处理成本`;

    const haikuCost  = cost(l2TotalIn, l2TotalOut, 'haiku') + cost(l3CtxIn, l3OutTok, 'haiku');
    const opusCost   = cost(l2TotalIn, l2TotalOut, 'opus')  + cost(l3CtxIn, l3OutTok,  'opus');
    const curCost    = l2Total + l3Total;
    if (selectedModel === 'haiku' || selectedModel === 'gpt4mini') {
      document.getElementById('tip-model').textContent = `模型选择: 已选最低成本方案，Opus将 ${fmt((opusCost - curCost)*30)}/月 更贵`;
    } else {
      document.getElementById('tip-model').textContent = `模型选择: 换用Haiku可节省 ${fmt((curCost - haikuCost)*30)}/月，适合简单分类任务`;
    }

    // Cache savings: Schema + System Prompt are reusable (Anthropic prompt caching 90% discount on cached tokens)
    const cacheableTok = (l2SchemaTok + l2SysTok) * 30;
    const cacheSaving  = (cacheableTok / 1e6) * MODELS[selectedModel].input * 0.90;
    document.getElementById('tip-cache-val').textContent = fmt(cacheSaving);

    const localEmbedSaving = embedTotal * 30;
    document.getElementById('tip-embed-val').textContent = fmt(localEmbedSaving);
  }

  // INIT
  update();
</script>
</body>
</html>

318 files · 7 not served here

Advantages

Enforces verification gates to prevent agents from making unverified claims. evidence: README.md
Provides a durable, structured knowledge base that compounds over time. evidence: README.md
Supports multi-agent orchestration with typed IPC and isolated write scopes. evidence: README.md

Limitations

Does not support dynamic or cyclic graphs; graph-engineering is bounded to static DAGs. evidence: README.md
Requires a specific transactional runtime for full functionality. evidence: AGENTS.md

Security

Risk level: Low risk

This brain runs

Agent instructions that run or fetch something (88)

AGENTS.md: V8.1 defines an explicit **5-Stage Worker Assembly Pipeline** in `workflows/worker-flows.md`, enforced by the canonical transactional runtime `python -m tools.worker_flow.cli`.
AGENTS.md: **loop-engineering** — Temporal-depth control through bounded Trigger → Execute → Verify → State loops, durable contracts, hard budgets, and stop/recovery rules.
AGENTS.md: `## Success Metrics` — minimum observable result for one successful run.
GEMINI.md: V8.1 defines a transactional **5-Stage Worker Assembly Pipeline** (`workflows/worker-flows.md`) enforced by `python -m tools.worker_flow.cli`.
skills/agent-teams-command/SKILL.md: <execute>
skills/agent-teams-command/SKILL.md: Run regression tests.
skills/agent-teams-command/SKILL.md: </execute>
skills/agentic-engineering/SKILL.md: Run the adoption gate: use an agent only when ambiguity/adaptation outweigh orchestration, verification, and maintenance cost.
skills/agentic-engineering/SKILL.md: <execute>
skills/agentic-engineering/SKILL.md: only host code may execute tools or decide continuation.
skills/agentic-engineering/SKILL.md: Run a thin loop: understand -> plan -> smallest change -> targeted test -> inspect diff/state -> broader check.
skills/agentic-engineering/SKILL.md: </execute>
skills/ai-six-sigma-property-os/SKILL.md: bounded agents execute and audit;
skills/ai-six-sigma-property-os/SKILL.md: <execute>
skills/ai-six-sigma-property-os/SKILL.md: </execute>
skills/anthropic-os/SKILL.md: Run Four-C in order: **Context** (truth/history), **Connections** (systems/accounts), **Capabilities** (skills/SOPs/evals), **Cadence** (triggers/reviews).
skills/anthropic-os/SKILL.md: <execute>
skills/anthropic-os/SKILL.md: Run dual prediction when useful: human and agent predict outcome independently;
skills/anthropic-os/SKILL.md: </execute>
skills/behavior-design/SKILL.md: <execute>
skills/behavior-design/SKILL.md: **H3 SOP:** write `trigger -> action -> evidence -> recovery -> stop` so a missed run has a next move.
skills/behavior-design/SKILL.md: </execute>
skills/behavior-design/SKILL.md: run a short observation period before optimizing.
skills/cognitive-compile/SKILL.md: <execute>
skills/cognitive-compile/SKILL.md: </execute>
skills/context-manager/SKILL.md: <execute>
skills/context-manager/SKILL.md: </execute>
skills/creativity-engine/SKILL.md: <execute>
skills/creativity-engine/SKILL.md: </execute>
skills/daily-okr/SKILL.md: Run one daily objective through seven linked Key Results.
skills/daily-okr/SKILL.md: <execute>
skills/daily-okr/SKILL.md: **Behavior:** execute one minimum behavior linked to the objective.
skills/daily-okr/SKILL.md: </execute>
skills/deep-research/SKILL.md: <execute>
skills/deep-research/SKILL.md: Run a broad discovery pass;
skills/deep-research/SKILL.md: Run a gap-fill pass only for decision-critical unknowns.
skills/deep-research/SKILL.md: Run an adversarial pass: strongest counterevidence, alternative mechanism, stale data, selection bias, and missing base rate.
skills/deep-research/SKILL.md: </execute>
skills/graph-engineering/SKILL.md: Run the admission gate before drawing a graph:
skills/graph-engineering/SKILL.md: Probe candidate independence with a small dry run.
skills/graph-engineering/SKILL.md: <execute>
skills/graph-engineering/SKILL.md: Write the JSON contract and run `scripts/validate_graph_contract.py <contract.json> --strict`.
skills/graph-engineering/SKILL.md: node success alone cannot certify graph success.
skills/graph-engineering/SKILL.md: </execute>
skills/graph-engineering/SKILL.md: run only independent ready nodes.
skills/harness-engineering/SKILL.md: Run the trace gate: the harness must be able to show what the agent saw, proposed, called, changed, and verified.
skills/harness-engineering/SKILL.md: <execute>
skills/harness-engineering/SKILL.md: The host decides whether to execute, continue, checkpoint, or escalate;
skills/harness-engineering/SKILL.md: </execute>
skills/harness-engineering/SKILL.md: Verify permissions with actual environment boundaries and run an independent end-to-end trace.
skills/harness-engineering/SKILL.md: A model emits plausible success text with a tool request: validate and execute the tool in host code, then verify its result;
skills/harness-engineering/SKILL.md: A maintenance run finds no eligible change: emit a verified `NO_OP` receipt and zero external outputs rather than creating notification or PR noise.
skills/harness-engineering/SKILL.md: The scheduler ran on time but the state store was stale: block mutation, mark the run failed, and recover from the last verified checkpoint.
skills/harness-engineering/SKILL.md: A Graph worker completes twice after lease expiry: deduplicate by node/run identity and release successors only from one verified transition.
skills/knowledge-ops/SKILL.md: <execute>
skills/knowledge-ops/SKILL.md: **Verify:** read after write, check links/source refs, rerun exact-identity search, and run representative retrieval queries.
skills/knowledge-ops/SKILL.md: Use idempotency keys for machine append operations: derive the key from operation, canonical source identity, target, and run/batch id;
skills/knowledge-ops/SKILL.md: </execute>
skills/knowledge-ops/SKILL.md: make no whole-vault claim.
skills/loop-engineering/SKILL.md: <skill_contract> <input>A repeatable task with inspectable state, finite budgets, permissions, and an independent verifier.</input> <output>A validated Trigger -> Execute -> Verify -> State contract plus resumable run receipts.</output> <done>The declared metric or stop condition is su …[+161 chars]
skills/loop-engineering/SKILL.md: **Goal:** run until a defined end state or cap.
skills/loop-engineering/SKILL.md: <execute>
skills/loop-engineering/SKILL.md: **Act:** execute one bounded action and capture artifact/diff/receipt.
skills/loop-engineering/SKILL.md: </execute>
skills/loop-engineering/SKILL.md: A scheduled job fired but produced no run receipt: status is triggered, not completed;
skills/project-flow-ops/SKILL.md: <execute>
skills/project-flow-ops/SKILL.md: Run the verification named in the definition of done.
skills/project-flow-ops/SKILL.md: </execute>
skills/session-learn/SKILL.md: <execute>
skills/session-learn/SKILL.md: **Pattern** — a reusable Trigger -> Execute -> Verify -> State sequence.
skills/session-learn/SKILL.md: </execute>
skills/startup-evaluation/SKILL.md: <execute>
skills/startup-evaluation/SKILL.md: </execute>
skills/verify-before-claim/SKILL.md: <execute>
skills/verify-before-claim/SKILL.md: Run the selected check after the final material change.
skills/verify-before-claim/SKILL.md: </execute>
skills/wiki-ingest/SKILL.md: record the identity query in the run ledger.
skills/wiki-ingest/SKILL.md: <execute>
skills/wiki-ingest/SKILL.md: Give every machine log entry an idempotency key derived from run/batch, operation, canonical source identity, and target;
skills/wiki-ingest/SKILL.md: **Verify:** run targeted link, source-ref, frontmatter, block-ref, empty-file, navigation, and post-write exact-identity checks over touched files.
skills/wiki-ingest/SKILL.md: </execute>
skills/wiki-lint/SKILL.md: make no content mutation unless separately authorized.
skills/wiki-lint/SKILL.md: <execute>
skills/wiki-lint/SKILL.md: Run twelve checks and preserve file-level evidence:
skills/wiki-lint/SKILL.md: </execute>
skills/wiki-lint/SKILL.md: If repair was authorized, run the relevant check again after each bounded repair.
skills/workflow-audit/SKILL.md: <execute> Extract a bounded list of distinct work candidates.
skills/workflow-audit/SKILL.md: </execute>

Install scripts (2)

install.ps1: pwsh install.ps1
install.sh: sh install.sh

Low risk: scripts-present (install.ps1)
Low risk: scripts-present (install.sh)

Install from source

7 of its files are not served here, so there is no ZIP. Mark393295827/third-brain-v7-skills at 5a64514

What it is

How it works

Requirements

Set it up

Usage

Documentation by Mark393295827

Third Brain V8.1 Skills

Quick Start

1. Clone & Install

2. Try a Skill

3. Next Steps

The Problem

V8.1 Operating Model

Core Skills (Start Here)

🧠 Knowledge & Verification

📅 Daily Loop

🎯 Behavior & Creativity

🔬 Research & Quality

📊 Context & Engineering

Complete Skill List

📥 Ingestion & Knowledge

🔄 Daily Workflow

🎨 Behavior & Creativity

🔬 Research & Quality

🔄 Learning & Flow

📊 Context

🏗️ Engineering

💼 Strategy & Operations

How These Skills Fit Together

Architecture & Design

Design Layers

Skill Adoption Path

Installation

For Codex CLI (Primary host/kernel)

For Cursor

For Windsurf

Compatibility adapters: Claude Code / Gemini CLI

Example Workflows

Wiki Structure

Tools & References

What This Is NOT

Philosophy

Related Work

Contributing

License

Transparency Note

Advantages

Limitations

Security

This brain runs

Attribution and licence

Install from source