zainfathoni/cara/review-verify

Claude Code and agent skill by zainfathoni/cara

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Verifies PR review concerns against code, targeted tests, and browser behavior. Use before replying, clearing, resolving, or publishing review feedback.

Install

npx vskill@latest install zainfathoni/cara/review-verify

SKILL.md

Review Verify

Verify review concerns independently. CI and deployment are supporting signals, not proof.

Concern Record

Every discovered concern gets one record, including concerns found only in a review summary and comments that are outdated, hidden by a squash, or no longer attached to the current diff:

  • source and stable identifier, plus visibility (current-user-pending, teammate-visible, or ambiguous)
  • exact concern
  • status: addressed, still-open, uncertain, or not-applicable
  • code mapping: relevant current paths and lines, or why no current code maps to it
  • test evidence: command and result, existing coverage inspected, or why a targeted test cannot establish the claim
  • browser evidence when user-visible behavior needs it: scenario and result, or why it was unnecessary or blocked
  • gap: remaining risk, missing access, unavailable artifact, or none

Use only this status vocabulary:

  • addressed: current evidence demonstrates the concern is fixed.
  • still-open: current evidence demonstrates the concern remains.
  • uncertain: evidence is missing, conflicting, or blocked.
  • not-applicable: the concern no longer applies; explain the changed premise rather than treating an outdated or squashed comment as automatically closed.

ready-to-resolve is not a concern status. Derive it only when status is addressed and the concern belongs to an identifiable, teammate-visible, unresolved public thread. Explicit user authorization is still required to resolve it.

Steps

  1. Inventory the review state. Read PR metadata, review bodies, inline comments, active thread state, commits, current head, CI, and deployment state. Include current-user pending reviews and recover referenced commits when squashing or outdated comments obscures context. Read REFERENCE.md for commands. Complete when every discovered inline, summary-only, outdated, and squashed concern has a concern record and source identifier.
  1. Map each concern to the current implementation. Inspect the referenced code and commit history, then locate the current code path that satisfies or violates the concern. Treat supplied URLs as entry points rather than scope boundaries. Complete when every record has an exact current code mapping or an explicit explanation of why none exists.
  1. Gather targeted evidence. Run the smallest non-E2E tests that directly exercise each testable claim; inspect existing coverage where execution is unavailable. Use manual browser verification when tests do not establish user-visible behavior. Read REFERENCE.md for command selection and REFERENCE.md for the browser procedure. Complete when every record contains test evidence, any required browser evidence, and an explicit gap (including `none`).
  1. Classify every concern. Apply exactly one authoritative status from the vocabulary above based on current code and direct evidence. Use CI and deployment only to support that evidence. Complete when every record has one status with a rationale and no concern is inferred closed merely because it is outdated, squashed, deployed, or green in CI.
  1. Report verification and route follow-up. Report all concern records, then list ready-to-resolve as a derived subset and state the explicit next action for requested replies, clearing, resolution, or publication. Use REFERENCE.md for a compact output shape. Complete when the report accounts for every discovered concern and names every evidence gap and follow-up owner.

Mutation and Publication Safety

Verification is read-only, including staging data. Use reversible browser interactions and obtain explicit approval before saving real data. Stop after reporting unless the user explicitly requests a follow-up mutation or publication. Route pending-artifact deletion to review-clear, teammate-visible resolution to team-review-resolve, and fixes or replies to review-address; use the relevant publication skill for submitting a pending review. Keep replies pending unless public visibility was explicitly requested. Never submit an existing pending review as a workaround for making a reply visible, and never resolve a thread whose concern is anything other than addressed or which fails public-thread eligibility.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/zainfathoni/cara/review-verify/badge)](https://verified-skill.com/skills/zainfathoni/cara/review-verify)

Something wrong with this skill or its page? Tell us.