Boost AI Coding Efficiency with Qoder Slash Commands: A Practical Guide

When using AI coding assistants like Qoder, simple queries such as “What are the new features in Java 25?” often trigger costly project file scans, web searches, and memory lookups, wasting time and tokens.

Solution: Slash Commands

Qoder supports slash commands (e.g., /java, /github) that bypass these searches and directly invoke the large language model for an answer.

Using /java prevents Qoder from scanning files or performing web searches, delivering a concise answer based solely on its internal knowledge.

Don't search project's files! Don't search web! Don't search memory!

The accompanying java.md file explicitly instructs Qoder not to search project files, the web, or memory, ensuring the model answers directly.

Other Command Examples

/github

– Provides GitHub‑related knowledge (e.g., creating SSH keys, GitHub Actions). The command file contains a similar “don’t search” directive but allows web search when information is insufficient. /llm – A minimal command that simply asks the LLM to answer the question without any tool activation.

Don't search project files! Don't search web! Don't search memory!

These commands are useful for quickly obtaining model answers without extra overhead.

Customizing Commands

Commands are defined in markdown files that can include rules, skills, and memory settings. By explicitly disabling tools like search_file or search_web, you control token usage and response speed.

For example, adding the line Don't search project's files to a command file prevents file scanning, while still allowing other tools if needed.

Practical Command List for Java Projects

/java

– Answers Java‑related questions. /spring – Provides information on Spring Framework, Spring Boot, Spring Data, and Spring Cloud. /github – Covers GitHub topics such as SSH keys and actions. /sql – Generates SQL statements based on the project’s database. /bash – Generates Bash scripts and command‑line snippets.

Specifying Programming Language

When generating code, Qoder determines the target language from the file name (e.g., User.java, user.mjs). In ask mode, explicitly stating the programming language (e.g., Programming language: Bash) ensures the model produces the correct syntax.

Similarly, using /bash in a command file declares the environment and installed tools, allowing Qoder to produce ready‑to‑run Bash commands.

Summary of Best Practices

Use noun‑based names for slash commands to make them intuitive.

Disable irrelevant tools in command files to speed up responses and save tokens.

Specify the programming language to guarantee the generated code matches your needs.

Provide clear context and environment details while adhering to the single‑responsibility principle for focused, accurate outputs.

By applying these techniques, you can make Qoder’s AI assistance faster, more precise, and token‑efficient.