Decide for Openclaw

A self-learning decision framework that enables AI agents to safely automate consequential project choices over time.

ivangdavila
v1.0.1
Mar 9, 2026
5
3.9k
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install decide

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install decide using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is Decide?

The Decide skill empowers your AI agent to transition from constant questioning to safe autonomy by learning your decision-making patterns. Unlike simple preference logging, this Openclaw Skills entry focuses on high-stakes branching choices—such as architectural shifts, vendor selections, and budget allocations—ensuring every autonomous move is backed by historical context and explicit human confirmation.

It bridges the gap between manual oversight and intelligent automation by building a durable memory of why certain choices were made. By systematically analyzing the components of a decision rather than relying on vague preferences, the agent builds a reliable knowledge base that scales with your project's complexity.

Decide Use Cases

  • Changing project architecture or selecting a new technology framework.
  • Evaluating vendor selections or third-party API integrations.
  • Managing long-term project direction and consequential workflow shifts.
  • Automating repetitive choices once a clear pattern is confirmed by the user.
  • Calibrating confidence levels for high-consequence technical branches to avoid silent failures.

How Decide Works

  1. The agent frames a consequential choice as a specific question supported by material components like budget, risk, and timeline.
  2. It gathers context from local memory, domain-specific overrides, and past decision logs stored in the decide directory.
  3. The system compares the current context against stored patterns to find a material match based on components, not just keyword similarity.
  4. If a match exists and confidence is high, the agent recommends the choice; otherwise, it defaults to a conservative "ask-first" approach.
  5. Once a human confirms a choice, the full rationale and outcome are logged to refine future Openclaw Skills performance.

Decide Setup

To begin using this skill, you must establish the local state directory to store your agent's learning progress.

# Create the decision storage directory and subfolders
mkdir -p ~/decide/domains/

# Initialize the memory and decision logs
touch ~/decide/memory.md ~/decide/decisions.md

# Sync the skill with your agent environment
clawhub sync decide

Decide Data Schema & Taxonomy

The Decide skill organizes data locally within the ~/decide/ directory to maintain privacy and accessibility for Openclaw Skills integrations.

File/Folder Description
memory.md Contains durable decision rules, approval boundaries, and confirmed defaults.
decisions.md A chronological log of major decisions, including question, components, and rationale.
domains/ Directory for domain-specific component models, overrides, and unique exceptions.
components.md Reference guide for factors that materially affect the decision-making lifecycle.

Decide Advanced Features

  • Pattern Promotion: Automatically identifies repeated decisions in similar contexts to propose new autonomous rules.
  • Component-Based Logic: Moves beyond simple text matching to analyze material factors like reversibility and maintenance cost.
  • Domain Overrides: Supports specific logic branches for different technical domains or client requirements.
  • Non-Destructive Workspace Routing: Integrates with AGENTS and SOUL files to ensure decision checks happen before major writes.
  • Safety-First Defaulting: Ensures that any unvalidated or high-stakes choice requires explicit human approval by default.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*