- Home
- Skills
- Code Quality & Review
- epicenter impact analysis
epicenter impact analysis
Map Python call graphs, find dead code, and analyze refactoring impact via an MCP-powered static analysis engine.
Free
Works with the AI Agents you already use
epicenter impact analysis
Example session with this skill installed
What is the impact of refactoring the BaseProcessor.validate method in this repo? Are there any untested callers?
- Read your context and instructions
- Compiled the epicenter impact analysis
I have analyzed the call graph for BaseProcessor.validate. The impact spans 3 modules, including one indirect caller that is currently untested.
Impact Analysis: BaseProcessor.validate
| Dependent Symbol | Type | Distance | Flags |
|---|---|---|---|
processors.UserProcessor.validate | Override | 1 | - |
api.endpoints.submit_data | Direct Call | 1 | [HIGH FAN-IN] |
tasks.background_worker.process_queue | Indirect Call | 2 | [UNTESTED] |
tests.test_processors.test_base_validation | Test Path | 1 | - |
Analysis Summary
High Risk: api.endpoints.submit_data is a high fan-in node. Changes to the validation logic may trigger widespread failures in the ingestion API.
Test Gap: tasks.background_worker.process_queue reaches this method through an indirect reference but has no associated test coverage. A regression here will not be caught by the CI suite.
Inheritance: UserProcessor overrides this method; ensure the signature change is reflected in the subclass to avoid TypeError.
Next steps
- Add a unit test in
tests/test_workers.pythat exercises theprocess_queuepath. - Update the signature in
processors/user.pyto match your intended changes. - Re-run
index_codebase()after the refactor to verify the new graph integrity.
Connects securely to your tools. The creator never sees your data.
About this skill
Refactoring Python code is often a guessing game of grep and hope. Epicenter eliminates the risk by providing a real call-graph analysis to see exactly what breaks before you touch a single line. It maps direct and indirect dependencies, identifies untested impact zones, and finds dead code using static analysis rather than simple text matching.
What it does
- Impact mapping traces direct and indirect callers across the codebase, including inherited overrides and typed variable resolutions.
- Dependency analysis lists every symbol a specific class or function relies on to help you understand unfamiliar code blocks.
- Coverage reachability identifies which tests actually reach a symbol, highlighting high-risk areas marked as
[UNTESTED]. - Dead code detection surfaces functions and classes that no entrypoint or test reaches, serving as candidates for safe removal.
- Change coupling uses historical git data to find symbols that frequently change together, even if they lack a direct code path.
How it works
- Initialize the index by calling
index_codebaseon your project root to build the initial call graph snapshot. - Search for symbols using
find_symbolto get the exact qualified name of the class or function you intend to modify. - Analyze the blast radius by running
impact_ofto see every dependent symbol, ranked by distance and fan-in. - Audit safety via
test_coverageanddead_codeto confirm if your changes are guarded by existing tests.
Frameworks & tools
This skill operates as an MCP server using epicenter-mcp. It is designed specifically for Python codebases and integrates with git for historical coupling analysis.
Why this beats prompting it yourself
Standard LLM prompts rely on the context window and basic text search, which miss indirect callers and complex inheritance. This skill uses a dedicated static analysis engine to provide a mathematically sound graph of your code, catching dependencies that grep ignores.
Use cases
- Safe refactoring to see every affected module before renaming a core utility function.
- Code cleanup to identify and remove truly unreachable classes and methods.
- Risk assessment to flag
[UNTESTED]dependents that require new manual test cases. - Onboarding to map out the dependencies of a complex class before attempting a feature addition.
Known limitations
Limited to Python. Cannot resolve dynamic dispatch via getattr, string-based attribute access, or unannotated duck typing. Results reflect the last indexed snapshot, not live unsaved edits.
Python only. Misses dynamic dispatch (getattr), unannotated indirection, and mid-function type reassignment. Requires CLI history mining for coupling data. Results depend on periodic re-indexing.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 19 days ago
- Free with an Agensi account
Requires