sakura-review

Review software changes against accepted outcomes, diffs, and verification evidence to decide pass or repair.

3|1|Updated Sep 9, 2026
One-click install
npx skills add https://github.com/v20227/using-sakura-dev --skill sakura-review-v20227
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: sakura-review
Source: https://github.com/v20227/using-sakura-dev/tree/main/skills/sakura-review
Command: npx skills add https://github.com/v20227/using-sakura-dev --skill sakura-review-v20227

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? It prevents teams from treating "code written" or "tests compiled" as done by requiring a structured review of the actual diff, current verification evidence, and accepted behavior before a workstream is closed. ## Core Features & Use Cases - Evidence-Based Acceptance: Compares the integrated result against the workstream's accepted outcome, regression risk, security, data integrity, and architecture ownership, then records PASS, FIX_REQUIRED, or BLOCKED. - Independent Review Delegation: For non-trivial workstreams, it dispatches a read-only subagent review so completion is not decided solely by the implementation context. - Blocking vs Non-Blocking Findings: Each finding gets a concrete location, evidence, impact, and smallest useful repair direction, separating blockers from optional improvements. - Use Case: Before closing a full-stack feature, run a review that requires current contract or integration evidence plus an end-to-end user path, catching a stale test result that no longer matches the reviewed implementation. ## Quick Start Ask the assistant to review the current change against the accepted outcome and verification evidence before closing the workstream.

Frequently Asked Questions about sakura-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review a code change before closing a workstream?▼

Confirm the review scope, inspect the actual diff and surrounding code, and compare the result against the accepted outcome and current verification evidence. Record PASS only when evidence is current; otherwise record FIX_REQUIRED or BLOCKED with concrete findings.

What evidence is needed to accept a full-stack feature?▼

A full-stack outcome requires current contract or integration evidence plus an end-to-end user path. Separate frontend and backend checks alone are insufficient, and stale test results must be re-verified after any review-driven change.

When should an independent reviewer be used for code review?▼

For non-trivial workstreams, an independent read-only subagent review should run so completion is not decided only by the implementation context. If independent review is unavailable, that limitation is recorded rather than silently skipped.

What makes a review finding blocking versus non-blocking?▼

Blocking findings prevent acceptance, such as missing integration evidence, regressions of accepted behavior, or unresolved license provenance issues. Non-blocking notes are improvements that do not affect the accepted outcome and can be handled in separate work.

Can a passing local test close an incomplete milestone?▼

No. A local passing check cannot close an incomplete milestone. Every required acceptance condition must be reconciled with current evidence, and unresolved required items stay open until the project task records final acceptance.