From hb-ai-toolkit
Reviews all files changed in the current branch vs a base branch. Auto-routes each file to the right reviewer, runs PR-level checks, logs recurring patterns, and outputs a verdict. Run before pushing or opening a PR.
How this skill is triggered — by the user, by Claude, or both
Slash command
/hb-ai-toolkit:review-pr [base-branch][base-branch]sonnetThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Review all files changed in this branch compared to `$ARGUMENTS` (default: `main`).
Review all files changed in this branch compared to $ARGUMENTS (default: main).
Base branch: use $ARGUMENTS if provided, otherwise main.
git log <base>...HEAD --oneline
Read commit messages to understand the change intent.
git diff <base>...HEAD --name-only
Get the list of changed files. Skip deleted files and binary files.
From the commit messages:
This shapes the review: bug fixes should have a regression test; refactors must not change observable behaviour; features must follow existing patterns.
For each changed file, read its content plus the diff:
git diff <base>...HEAD -- <file>
Apply ${CLAUDE_PLUGIN_ROOT}/claude/agents/code-reviewer.md to every file.
Weight findings toward changed lines — don't flag pre-existing issues unless they are directly relevant to the change.
For bug fixes:
warning.For refactors:
warning — these are breaking changes.For features:
For all change types:
console.log / console.error left in non-test production code.TODO / FIXME comments added in this diff.After reviewing all files, pass the full review output to ${CLAUDE_PLUGIN_ROOT}/claude/agents/issue-logger.md in Mode 1.
This will update claude/rules/recurring-issues.md if any issue class appeared 3+ times in this run.
## PR Review
**Base:** <base>...HEAD
**Change type:** {Bug fix | Feature | Refactor | Chore}
**Files reviewed:** {n}
---
{Individual review for each file using the code-reviewer output format}
---
## PR-Level Notes
### Must fix
- {critical cross-cutting issues}
### Suggestions
- {warnings spanning multiple files}
---
## Verdict
{APPROVE | REQUEST CHANGES | NEEDS DISCUSSION}
**Rationale:** {one or two sentences}
Verdict guide:
APPROVE — no critical issues; warnings are minor or already acknowledgedREQUEST CHANGES — one or more critical issues, or a missing regression test on a bug fixNEEDS DISCUSSION — the approach is questionable but you're uncertain — flag for team conversationnpx claudepluginhub hackberrybay/hb-ai-toolkit --plugin hb-ai-toolkitProvides CDSS development patterns for drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), and alert classification integrated into EMR workflows.