Code intelligence for AI agents — semantic search, symbol lookup, call graphs, dependency mapping, blast radius analysis, and ownership resolution via mcp-code-search.
| Workflow | Tool Calls | What It Achieves |
|---|---|---|
| Find Code | 1 | Semantic search across repos |
| Understand Symbol | 2-3 | Definition + call graph + owners |
| Blast Radius | 2-3 | Impact analysis before changes |
| Build Context | 1 | All files needed for a task |
- Grep-based search misses semantic matches
- No understanding of who calls what
- Changes made without knowing blast radius
- Reviews sent to wrong people (no ownership data)
- Natural language code search ("payment retry logic")
- Full call graph tracing (callers + callees)
- Blast radius calculated before any change
- Code owners identified for proper review routing
git clone https://github.com/zavora-ai/skill-code-search-navigation.git \
~/.skills/skills/code-search-navigationRequired: mcp-code-search (8 tools)
Cross-MCP:
mcp-github— find code → create PRmcp-jira— impact analysis → create issue
code-search-navigation/
├── SKILL.md # Decision tree + workflows
├── scripts/
│ └── blast_radius.py # Impact calculator from call graph data
├── references/
│ ├── tool-sequences.md # 8 tools
│ ├── cross-mcp-workflows.md # Code Search + GitHub + Jira
│ └── examples.md # Find, understand, blast radius
├── README.md
└── LICENSE
User: "Where do we handle payment retries?"
Result:
Found: src/retry.rs:23 — pub fn retry_with_backoff(...)
Callers: checkout.rs:12, webhook.rs:30
Owner: @payments-team
python scripts/blast_radius.py '{"callers": ["a.rs", "b.rs"], "dependents": ["mod_x", "mod_y"]}'
# → {"risk_level": "medium", "recommendation": "Notify code owners. Run full test suite."}| Metric | Target |
|---|---|
| Trigger rate | 90% on code navigation queries |
| Context quality | Always include file path + line numbers |
| Blast radius | Calculated before recommending changes |
![]() James Karanja Maina |
|---|
Apache-2.0 — Part of ADK-Rust Enterprise. Built with ❤️ by Zavora AI
