| name | using-evocli |
| description | Use this skill when working in an EvoCLI-powered project to leverage its unique capabilities: long-term memory, skills, workflow automation, and code intelligence. |
Using EvoCLI Effectively
EvoCLI transforms a standard repository into an intelligent, self-documenting workspace. By using EvoCLI tools instead of raw shell commands, you ensure that your actions are recorded, analyzed, and optimized over time.
The EvoCLI Workflow
When you enter an EvoCLI project, your first step should always be context retrieval.
- Recall Context: Use
memory.query to see what has been done recently and what constraints are in place.
- Check Skills: Run
skill.list to see if there are existing automations for your current task.
- Execute with Tools: Use the specialized
fs.*, git.*, and code_intel.* tools. These tools are faster than shell equivalents and provide structured feedback to the LLM.
- Contribute to Memory: When you make a decision or discover a pattern, use
memory.add to store it for future sessions.
Memory Integration
EvoCLI's memory system is its greatest asset. It prevents "context drift" where the AI forgets project rules over long sessions.
- Constraint Memories: Use these for architectural rules (e.g., "Always use functional components").
- Episodic Memories: These are automatic logs of what you did. You can query them to find "how did I fix that bug last week?".
- Semantic Memories: These store the "why" behind the code.
Skill Execution
Skills are more than just scripts; they are verified workflows.
- Drafting: If you find yourself performing a complex sequence of steps, use
skill.draft.
- Verification: Always run
skill.verify before trusting a new skill. This runs a multi-level dry-run to ensure safety.
- Trusted Skills: Once a skill is trusted, it can be invoked as a single command, drastically reducing token usage and execution time.
Code Intelligence
EvoCLI's code_intel tools use the Rust Host to parse your codebase.
- Symbol Search: Find definitions and references across the entire project instantly.
- Dependency Mapping: Understand how a change in one file affects the rest of the system.
- AST Grep: Perform structural searches that are more accurate than regex.
Best Practices
- Prefer Tools over Shell:
fs.read is better than cat because it handles large files gracefully and provides line numbers.
- Be Explicit in Memory: When adding to memory, use clear, concise language.
- Leverage the TUI: The
evocli tui provides a rich environment for debugging complex tool chains.
- Evolution is Key: Pay attention to EvoCLI's suggestions for new skills. The system learns from your specific coding style.
Security and Safety
EvoCLI operates on a "Safety First" principle.
- All shell commands are checked against a whitelist.
- Destructive operations (like
git.reset --hard) require user confirmation unless the skill is explicitly "Trusted".
- The Python environment is isolated, preventing prompt injection from accessing your system directly.
