code-review
Code Review
Overview
Perform a rigorous, adversarial review that cross-checks story claims and acceptance criteria against the real code changes. Prioritize concrete, reproducible findings with file and line references. If the code is clean, explicitly say so.
Quick Start
- Determine scope: story file + repository, or a diff/PR.
- Identify repo root (use git) and collect changed files via
git status --porcelain,git diff --name-only, andgit diff --cached --name-only. - Load story/spec inputs if provided; otherwise ask for the story/spec or acceptance criteria.
- Exclude non-application files and folders (
.cursor/,.windsurf/,.claude/, and other tooling/config folders unless explicitly requested).
Workflow
Follow the detailed workflow in references/workflow.md. Use these guardrails when adapting it to a standalone repo:
- If
{project-root}/workflow-config.yamldoes not exist, treat config values as optional and ask the user for missing items that affect output (language, skill level, story path, etc.). - If planning artifacts (architecture/ux/epics) are unavailable, proceed with the story/spec and code changes only; note missing context explicitly.
- If no story file exists, replace “story validation” with “requirements validation” using the user-provided criteria or a PR description.
Review Expectations
- Validate each Acceptance Criterion and each task marked complete. Missing ACs are HIGH severity; tasks marked complete but not implemented are CRITICAL.
- Cross-check story File List against actual git changes; report discrepancies.
- Cover security, performance, error handling, test quality, and maintainability.
- If no issues remain after a second pass, report a clean review.
Output
- Present findings grouped by severity with evidence (
file:line), and a short impact statement. - If asked, fix issues or create action items; otherwise provide a concise remediation list.
- When applicable, update story status and sprint status in accordance with the workflow.
References
references/workflow.mdfor the full standalone code review flow.references/discover-inputs.mdfor input discovery/loading strategy.references/checklist.mdfor the review completion checklist.
More from sungkhum/agent-skills-pack
edge-case-hunter
Exhaustive edge-case review that traces branching paths and boundary conditions in a diff or code sample and reports only unhandled cases. Use when the user asks to hunt edge cases, boundary conditions, missing guards, or wants an edge-case-focused review.
11market-research
Standalone market research workflow. Use when the user asks for market research, competitive landscape analysis, customer needs, market trends, or a market research report about a business idea.
10domain-research
Standalone domain or industry research workflow. Use when the user asks for domain research, industry analysis, sector overviews, regulatory landscape, or a domain research report.
10tech-writer
Standalone technical writer agent. Use when the user asks to write or update technical documentation, create Mermaid diagrams, explain complex concepts, validate docs against standards, or generate comprehensive project documentation.
10ux-designer
Standalone UX designer agent. Use when the user asks for UX design, user research, interaction design, UI patterns, experience strategy, or guidance creating a UX design specification.
9create-ux-design
Standalone UX design specification workflow. Use when the user asks to create UX design specs, plan UX patterns, define UX strategy, or produce a UX design specification document.
9