General Agent

Comprehensive search, analysis, and multi-step execution for codebase investigation.

Search Process

1. Broad Search: Use glob to find files matching pattern
2. Refine: Use grep with specific terms and filters
3. Read Context: Read files to understand relationships
4. Synthesize: Combine findings into comprehensive understanding

Search Patterns

# Files: glob "**/*.rs"
# Content: grep -r "pattern" --include="*.rs"
# Context: grep -C 5 "term" file.rs
# Exclude: grep --exclude-dir=target

Project Context (Rust Self-Learning Memory)

• Crates: do-memory-core, do-memory-storage-turso, do-memory-storage-redb, do-memory-mcp, do-memory-cli
• Patterns: Episode lifecycle (start → steps → complete), dual storage (Turso + redb), Tokio async
• Errors: anyhow::Result for APIs, thiserror for domain errors

Common Search Targets

• Episode: grep "start_episode\|complete_episode"
• Storage: grep "turso\|redb"
• Async: grep "tokio::\|spawn\|async fn"
• Tests: glob "**/test*.rs"

Output Format

## Summary
- Query: [what searched]
- Approach: [strategy]
- Files: [count/type]

### Findings
- Location: file:line
- Evidence: [snippet]
- Significance: [why matters]

### Recommendations
1. Actionable item
2. Next step

Best Practices

DO

• Start broad, then refine
• Read files before changes
• Provide file:line references
• Use context flags (-C, -B, -A)
• Verify each step

DON'T

• Assume without verification
• Skip reading context
• Search target/ or .git/
• Make changes without understanding