A strategic error-correction skill that inserts breakpoint self-critique at risky inflection points to prevent downstream mistake propagation.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install self-criticism
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install self-criticism 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 Self-Criticism?
Self-Criticism is a specialized skill designed for AI agents to improve output quality by introducing intentional pauses at critical decision-making junctions. By inserting a short critique before branching work, after encountering user friction, or during risky handoffs, agents can identify and rectify flawed assumptions before they multiply into complex downstream errors.
Built to integrate seamlessly with SOUL and AGENTS architectures, this skill focuses on mid-task interruption rather than background review. It ensures that the AI remains outcome-focused, using the smallest set of hard questions necessary to validate a plan, thereby maintaining momentum while drastically reducing the need for late-stage rework within the Openclaw Skills ecosystem.
Self-Criticism Use Cases
- Preventing the propagation of wrong assumptions during complex code refactoring or system architecture design.
- Validating high-level project plans and task decompositions before execution begins.
- Adjusting agent behavior after user feedback suggests a drift in process or objective.
- Performing a safety check before committing irreversible edits or external actions.
How Self-Criticism Works
- The agent identifies a natural inflection point, such as a major recommendation or a subtask decomposition.
- A breakpoint is triggered based on local rules stored in memory.md.
- The skill evaluates the necessary critique depth (Light, Standard, or Deep) based on the task's risk and blast radius.
- A focused self-critique is performed using templates from checkpoints.md.
- If framing errors are found, the agent resets the workspace frame and revises the plan before continuing.
- Durable lessons from the interaction are saved to incidents.md to refine future trigger points.
Self-Criticism Setup
To initialize the self-criticism environment and local memory storage, run the following command:
# Navigate to your skill directory and run setup
clawhub install self-criticism
# If ~/self-criticism/ does not exist, initialize it
cat setup.md | openclaw run
Self-Criticism Data Schema & Taxonomy
The skill maintains a local directory at ~/self-criticism/ with the following structure:
| File | Description |
|---|---|
memory.md |
Durable trigger maps, depth preferences, and integration status. |
checkpoints.md |
Named breakpoint prompts organized by workflow phase. |
incidents.md |
Log of late catches and missed opportunities to improve insertion points. |
archive/ |
Directory for resolved incidents and retired critique checkpoints. |
Self-Criticism Advanced Features
- Multi-level depth logic that scales from light routine checks to deep architectural scrutiny.
- Dynamic trigger memory that learns from user rework to improve future breakpoint accuracy.
- Workspace steering for SOUL and AGENTS to embed the critique habit directly into the agent's operating logic.
- Local-first privacy model ensuring all critique logs and incident reports stay on your machine.
SKILL.md
Loading
Related Openclaw Skills
Careers Page Generator
A specialized tool for AI agents to design, audit, and optimize high-converting careers pages and employer branding content.
Face
Give your AI agents a professional human appearance with fast avatar shortlists and transparent background options.
Save Token
A specialized guiding skill that empowers AI agents to minimize token consumption by identifying and eliminating redundant conversation context.
AIPPT: Automated Presentation Generator
AIPPT is an automated skill that generates professional PowerPoint presentations from text descriptions or uploaded documents including docx and images.
Gemini Image Gen
A high-performance AI image generation tool utilizing Google Gemini models for text-to-image and reference-guided design.
MarkItDown Document Converter
A high-performance document conversion tool that leverages Microsoft's MarkItDown library to turn various file formats into structured Markdown.
Featured*
