About this skill

Problem It Addresses

Code review often depends on individual experience and a single pass through a diff, which can miss architecture boundaries, performance hot spots, security flaws, test impact, and product-requirement drift. Large MRs mix high-risk and low-risk files, and judging from diff fragments alone can produce false positives or false negatives. eagle-code-reviewer targets Git repository MRs/PRs and turns review into a traceable, role-based workflow.

How It Works and Where It Fits

The skill collects the MR/PR URL, optional PRD, and Feishu report folder, then creates a Feishu document after core tool checks for git, glab, and graphify pass. It reads the diff, states the change intent, and applies batching, tiering, or sampling based on file count and diff size: core business logic, auth, database, and API changes receive deep review, while config, tests, and generated code receive structural scans. Review roles include architect, performance engineer, security engineer, senior developer, and test engineer; a product manager role is added when a PRD is provided. The test engineer must use graphify call chains to assess impact and list P0/P1/P2 test cases. Every finding must be grounded in full function context, callers, dependencies, or data flow, not just the diff snippet, with language-specific checks for Java, Go, and Python.

It fits team PR reviews that need a stable template and auditable reports. Missing Feishu folder or core tools can stop execution, and oversized diffs fall back to sampling with a split-MR warning.

Screenshots

Use Cases

Before a GitLab MR is merged, use architecture, security, and QA roles to create a Feishu review report.
For large MRs, apply tiered or sampled review to core Tier 1 logic and test files.
When a PRD is available, let the product manager role check feature completeness and edge cases.
Use graphify call chains to identify impact scope and output P0/P1 test cases.

Best For

Senior engineers gatekeeping GitLab MR merges need stable architecture and security review output.
Engineers aligning PRDs with code need checks for missing features and edge cases.
QA engineers assessing regression scope need call-chain evidence for P0/P1 test cases.
Engineering teams archiving review conclusions in Feishu need batched results written to docs.