AI-Agent Resources

ARCHIVED — This repository is read-only. It is the preserved workflow toolkit that powered a self-organizing AI agent built for the Antigravity platform (December 2025). The workflows are no longer maintained, but the kit is kept here as a working reference for building structured AI-agent knowledge and planning systems.

A collection of 27 markdown prompt-workflows and 4 Python automation scripts that give an AI agent an "operating system" for long-running, autonomous work — a persistent knowledge base ('Brain'), session rituals, and a full product-planning pipeline.

Live preview: 🌐 https://ai-agent-five-psi.vercel.app

Content-Type: text/plain + Content-Disposition: inline (set in vercel.json) lets the .py and .md assets render in the browser instead of downloading.

---

Table of Contents

• What is this?
• Use cases
• Key features
• How it works
• Repository layout
• Core workflows
• Planning workflows
• Python scripts
• Setup & installation
• Development
• Status & limitations
• License

---

What is this?

This repo bundles the building blocks of a self-organizing AI agent:

• Prompt-workflows (.md) that instruct an AI agent to follow disciplined rituals — naming conventions, folder structure, and operational procedures — so it can run autonomously for days without losing context or drifting.
• Python automation that keeps the agent's workspace healthy: indexing, document health checks, and failure-pattern analysis, offloading repetitive work from the LLM.

It was originally built for Antigravity (an agent platform) in December 2025 as AI_AGENT_INSTALL.md — a one-file "installer" that tells an agent how to scaffold the whole system in any workspace.

Use cases

• Long-running agent sessions — give an AI agent a persistent memory (Topic, Plan, Knowledge) so it can pick up where it left off across sessions.
• Autonomous task tracking — structured, machine-readable artifacts (TOPIC_001_..., PLAN_001_...) instead of free-form chat output.
• Failure learning — log failures, let the agent detect patterns and improve.
• Product & project planning — a repeatable 13-workflow planning phase that compiles into a single Planning Blueprint.
• Workspace maintenance — Python scripts that audit and auto-index a growing markdown knowledge base.

Key features

• 📇 Named & versioned artifacts — TOPIC_NNN, PLAN_NNN, FIND_NNN, K_NNN with a standard slug format.
• 🧠 Session-start ritual (wake_up.md) — loads the 'Brain' (Topic) and 'Rencana' (Plan) into working memory.
• 🔁 Discussion loop (discussion_cycle.md) — discuss and record a summary back into the Topic file.
• 🔬 Deep research (deep_research.md) — online research via MCP tools.
• 📊 Full planning phase — 13 planning workflows converging into one Planning Blueprint.
• 🤖 Python tooling — workspace analyzer, auto index updater, document health analyzer, failure analyzer.

How it works

Session lifecycle

flowchart TD
    A["🚀 wake_up.md — Session start ritual"] --> B{"Agent works"}
    B --> C["🔬 deep_research — research via MCP"]
    B --> D["💬 discussion_cycle — discuss & write summary"]
    B --> E["⚠️ log_failure — record tool/command failures"]
    C --> F["📚 create_knowledge — persist insight to KB"]
    D --> F
    E --> G["🐍 failure_analyzer.py — detect failure patterns"]
    B --> H["create_topic / create_plan / create_finding"]
    H --> I["🔎 self_audit — data integrity & strategy check"]

Loading

Planning pipeline

flowchart LR
    A["✅ validate_idea\nGO / PIVOT / KILL"] --> B["💡 ideate_project\nDesign Thinking"]
    B --> C["🧠 brainstorm_session\n8 techniques"]
    C --> D["💰 cost_benefit"]
    C --> E["🔬 feasibility_study\nTELOS"]
    D --> F["🎯 create_lean_canvas"]
    E --> F
    F --> G["🔲 define_scope / 🔧 tech_stack_eval"]
    G --> H["👥 raci_matrix + risk_register"]
    H --> I["📜 create_charter"]
    I --> J["📋 compile_blueprint\n→ 1 master Planning Blueprint"]

Loading

Repository layout

AI_AGENT_INSTALL.md           # Installer: instructs an AI agent how to scaffold the system
AGENTS.md                     # Rules for AI agents editing this repo (anti-hallucination)
index.html                    # Landing page (Vite + static assets)
vercel.json                   # Vercel headers: render .py/.md inline
.github/workflows/build.yml   # CI: builds the site on every push
public/
├── scripts/                  # Python automation scripts (served statically)
└── workflows/                # Markdown prompt-workflows (served statically)
    ├── index.md              # Workflow overview / table of contents
    ├── wake_up.md …          # Core workflows
    └── planning/             # Planning & product-thinking workflows

Core workflows

File	Description
index.md	Workflow overview / table of contents
wake_up.md	Session-start ritual — loads 'Brain' (Topic) and 'Rencana' (Plan) into working memory
create_topic.md	Create a new Topic in the AI-Agent system
create_plan.md	Create a new Plan (implementation plan)
create_finding.md	Create a new Finding (bug / issue)
create_knowledge.md	Create a Knowledge entry in the Knowledge Base
deep_research.md	Perform deep online research on a topic using MCP tools
discussion_cycle.md	Discuss a topic and record the summary into the Topic file
log_failure.md	Record tool/command failures to the failure log for learning
self_audit.md	Autonomous introspection for data integrity & strategy alignment

Planning workflows

File	Description
brainstorm_session.md	Guided brainstorming — 8 techniques (Mind Map, SCAMPER, Crazy 8s, Brainwriting, …)
validate_idea.md	Full validation checklist — GO / PIVOT / KILL decision framework
ideate_project.md	Project ideation using Design Thinking (Empathize → Define → Ideate → Prototype → Test)
customer_interview.md	Guided customer interview with The Mom Test framework
cost_benefit.md	Cost-Benefit Analysis — ROI calculation, payback period, financial viability
feasibility_study.md	TELOS feasibility — Technical, Economic, Legal, Operational, Schedule
create_lean_canvas.md	One-page business model (Lean) — 9 blocks (Problem, Solution, UVP, …)
define_scope.md	Explicit In/Out of scope definition to prevent scope creep
tech_stack_eval.md	Technology comparison with weighted scoring (Frontend / Backend / DB)
raci_matrix.md	RACI matrix — Responsible / Accountable / Consulted / Informed per task
risk_register.md	Risk identification, scoring & mitigation (probability × impact matrix)
create_charter.md	Project charter — objectives, scope, stakeholders, timeline, budget
compile_blueprint.md	Compile outputs of the 12 planning workflows into one master "Planning Blueprint"

Python scripts

Script	What it does
analyze_workspace.py	Scans and indexes all markdown files in the workspace → workspace_index.json
auto_index_updater.py	Auto-updates index.md in each folder (Topic, Plan, Find, Knowledge, Research) and reports changes
document_health_analyzer.py	Health checks: broken-link detection, orphans, index validation, cross-reference map, health score
failure_analyzer.py	Parses failures.md, groups failures (Tool + Error Type), detects new patterns, generates a report

Setup & installation

⚠️ Targets the Antigravity agent platform and is not maintained. Keep for reference.

To scaffold the entire system in a fresh workspace with any compliant AI agent:

1. Drop AI_AGENT_INSTALL.md into the project and mention @AI_AGENT_INSTALL.md install.
2. The agent creates the folder structure and .agent/STANDARDS.md (naming, folder layout, operational procedures).
3. Use wake_up.md at the start of each session to load context into working memory.

Development

bun install
bun run dev        # local dev server
bun run build      # production build → dist/
bun run preview    # preview the production build

Static assets live in public/ and are copied to dist/ at build time. A GitHub Actions workflow (.github/workflows/build.yml) runs bun install + bun run build on every push.

Status & limitations

• Status: Archived, read-only (last updated December 2025).
• Built specifically for the Antigravity platform — not a general-purpose framework.
• The workflow documents are in Indonesian; no English translation is maintained.
• No automated tests or runtime; scripts are standalone Python utilities.

License

MIT © 2025 mifdlaldev