A precision-engineered system prompt installer that strips corporate filler and verbosity from AI responses across English and Chinese.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install talk-normal
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 talk-normal 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 talk-normal?
talk-normal is a specialized behavior modifier designed to solve the problem of LLM slop—the wordy, repetitive, and overly corporate tone often found in modern AI responses. By injecting a set of highly refined rules into your environment, it forces the model to communicate with technical brevity. As a core part of the Openclaw Skills ecosystem, it targets specific rhetorical tics and structural inefficiencies that bloat token usage and degrade the user experience.
Unlike traditional prompt engineering that requires manual effort for every interaction, this skill functions as an always-on filter. It has been benchmarked to provide a character reduction of 56% to 71% across leading models like GPT-5.4 and GPT-4o-mini, ensuring that the information density of your agent remains high without losing critical context in either English or Chinese.
talk-normal Use Cases
- Reducing token costs and latency by cutting unnecessary conversational filler.
- Forcing AI agents to lead with direct answers rather than restating the user's question.
- Cleaning up bilingual (English/Chinese) technical documentation and chat outputs.
- Standardizing agent behavior across a team to ensure a professional, non-verbose tone.
- Removing repetitive closing phrases and hypothetical follow-up menus from automated workflows.
How talk-normal Works
- The user executes the installation script within their Openclaw workspace.
- The skill identifies the AGENTS.md file and checks for existing markers.
- It injects or updates a curated block of system rules from prompt.md between specific BEGIN and END markers.
- On every subsequent interaction, the AI agent reads these persistent instructions.
- The agent applies structural discipline, removing banned opening phrases, hedging, and rhetorical slop before generating the final response.
talk-normal Setup
To integrate this behavior into your workspace using Openclaw Skills, run the following commands:
# Install the skill to your AGENTS.md
bash install.sh
# To update to the latest anti-slop rules
clawhub update talk-normal
bash install.sh
# To remove the rules and revert to default behavior
bash install.sh --uninstall
talk-normal Data Schema & Taxonomy
The skill manages persistent configuration data through the following file structure:
| File | Role |
|---|---|
AGENTS.md |
The primary destination where the system prompt rules are injected and stored. |
prompt.md |
Contains the raw source of the curated anti-slop rules used for injection. |
TEST_RESULTS.md |
Holds benchmark data documenting character reduction and information preservation. |
Rules are categorized into functional groups such as Filler/Hedging, Structural Discipline, Closing Patterns, and Rhetorical Tics.
talk-normal Advanced Features
- Idempotent installation: Re-running the installer safely replaces existing rules without duplicating content or corrupting your AGENTS.md.
- Bilingual Support: Specific rulesets optimized for both English and Chinese linguistic patterns and common AI 'translation' slop.
- Model Agnostic: Compatible with any LLM that honors system instructions, including Claude, Gemini, GPT, and DeepSeek.
- Targeted Slop Eradication: Each rule update is mapped to specific corporate phrasing patterns caught in the wild.
SKILL.md
Loading
Related Openclaw Skills
Replication Strategy Selector
A specialized architectural tool for selecting and configuring optimal database replication topologies based on availability, consistency, and latency requirements.
Partitioning Strategy Advisor
A technical advisor for selecting sharding strategies, partition keys, and secondary indexing methods to ensure horizontal scalability in distributed databases.
Storage Engine Selector
A 7-dimensional trade-off analysis tool for selecting the optimal database storage engine architecture based on specific workload patterns.
Academic Research
An automated research assistant that retrieves, filters, and synthesizes academic literature into structured reports and presentations.
Replication Failure Analyzer
A professional diagnostic framework for mapping database replication symptoms to failure classes and generating structured remediation plans.
AI Sentinel
An automated intelligence system that collects, analyzes, and reports the latest developments in the AI field from multiple global sources.
Featured*
