Agentic AI CodeSentinel-Code Review, Dependency & Security analysis

Software teams today face an ever-growing backlog of code reviews, security audits, and compliance checks. Manual code review is slow, inconsistent, and often misses critical security vulnerabilities or architectural anti-patterns. CodeSentinel is an Agentic AI intelligent code review platform that leverages Amazon Bedrock to automate the entire code review lifecycle--from code ingestion through dependency analysis, individual file review, and final report generation--in a single, orchestrated workflow.

Unlike traditional static analysis tools, CodeSentinel uses a multi-step agentic workflow with extended thinking capabilities, orchestrating code review through five steps:

Step 1 - Load Code: Ingests source code from CodeCommit branches, CodeCommit PRs, S3 URIs, ZIP uploads, AWS Lambda functions, and Bedrock Agents. Supports .py, .js, .ts, .java, .go, .rb, .php, .c, .cpp, .h, .cs, .rs, .swift, .kt, .sh, .yaml, .yml, .json, .tf, and .sql files.

Step 2 - Build Dependency Graph: Parses import/require/from statements to build a file-to-file dependency graph. Performs topological sort for bottom-up review order, detects circular dependencies, identifies AWS services used per file, extracts environment variable references, and lists external libraries.

Step 3 - Review Files with Fix Recommendations: Reads each file in dependency order, passing a rolling summary of prior findings as context. Analyzes for Security (OWASP Top 10, hardcoded secrets, injection risks), AWS Well-Architected Framework (all 6 pillars), Bugs and Logic Errors, Code Quality, Performance, and Cross-File Issues. Each finding includes severity (CRITICAL/HIGH/MEDIUM/LOW/INFO), file name, approximate line number, category, issue description, impact, and fix recommendation.

Step 4 - Generate Report: Compiles all findings into a structured Markdown report with severity summary. Automatically exports the report and persists full metadata for historical tracking, trend analysis, and delta reviews.

Step 5 - Next Review: Loads the previous review's findings, takes the latest code version, and identifies which issues are fixed, which remain unresolved, and any new issues introduced.

Key Features:

• Multi-Source Code Ingestion from CodeCommit, S3, ZIP upload, AWS Lambda, and Bedrock Agent configurations
• Automatic Dependency Graph construction ensuring foundational modules are reviewed before dependents
• Six-Dimension Analysis covering Security, AWS Well-Architected, Bugs, Code Quality, Performance, and Cross-File Issues
• Severity-Ranked Findings with file name, line number, category, business impact, and fix recommendations
• Persistent Review History enabling trend analysis and audit trails
• Follow-Up Reviews that identify fixed, unresolved, and newly introduced issues

Target Audience: Software engineering teams, DevSecOps teams, cloud architects, and platform engineering teams at enterprises and ISVs who deploy workloads on AWS and need automated, consistent, and audit-ready code review at scale.