Reviews PR changes against the broader codebase to find inconsistencies, missed propagation, and stale references beyond the diff. Replicates senior reviewer codebase knowledge. Activates when reviewing PRs, checking consistency, or when changes touch shared entities across multiple files.
From pwnpx claudepluginhub ken2403/claude-paralell-dev-plugin --plugin pwThis skill is limited to using the following tools:
references/propagation-paths.mdSearches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Searches prompts.chat for AI prompt templates by keyword or category, retrieves by ID with variable handling, and improves prompts via AI. Use for discovering or enhancing prompts.
Implements structured self-debugging workflow for AI agent failures: capture errors, diagnose patterns like loops or context overflow, apply contained recoveries, and generate introspection reports.
A senior reviewer has the full codebase in their head. When they read a diff, they instantly know what other files should also change. This skill replicates that by building codebase context around the diff and checking for inconsistencies.
Apply on every PR review. The value scales with PR size and cross-cutting concerns.
Copy this checklist and track progress:
Codebase Consistency Review:
- [ ] Step 1: Build codebase context
- [ ] Step 2: Run local build
- [ ] Step 3: Trace change propagation
- [ ] Step 4: Check related files outside the diff
- [ ] Step 5: Verify cross-layer consistency
- [ ] Step 6: Compile findings
This is the most critical step. The usage graph is what a senior reviewer holds in their head.
Run the project's build, lint, and typecheck commands to collect concrete errors. Detect the appropriate commands from the project structure (Makefile, package.json, pyproject.toml, Cargo.toml, go.mod, etc.).
Build errors provide a mechanical safety net that catches every compile-time inconsistency that code reading alone might miss. Categorize errors by file and type.
For each changed entity from Step 1, verify the change has propagated to ALL dependent code:
For detailed propagation paths by change type, see references/propagation-paths.md.
For each candidate file identified in Step 1 (in usage graph but not in diff):
Also check for files that are in the diff but only partially updated — for example, a variable was renamed but string literals or form field references using the old name were not.
Check that changes are consistent across all layers of the application:
Output structured findings:
PR_INTENT: [one-line summary of what the PR is trying to accomplish]
CHANGED_ENTITIES: [list of entities modified by the diff]
BUILD_ERRORS: [categorized list or "none"]
MISSED_FILES: [file:line list with what needs updating, or "none"]
PARTIAL_UPDATES: [file:line list with what was missed, or "none"]
CROSS_LAYER_ISSUES: [list or "none"]
MIGRATION_ORDERING: [conflicts or "none"]
For each finding: