taste-review

Delegates ambiguous design and wording decisions to the Claude CLI for judgment calls.

3|Updated Nov 8, 2014
One-click install
npx skills add https://github.com/mintuz/.dotfiles --skill taste-review-mintuz
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: taste-review
Source: https://github.com/mintuz/.dotfiles/tree/main/agents/.agents/skills/taste-review
Command: npx skills add https://github.com/mintuz/.dotfiles --skill taste-review-mintuz

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When you hit a fuzzy decision — UI polish, prose phrasing, naming, or formatting — guessing leads to inconsistent results. This Skill shells out to the claude CLI to get a reasoned judgment call, then applies the recommendation. ## Core Features & Use Cases - Taste-Driven Decisions: Ask Claude to weigh options on ambiguous questions like naming, phrasing, or visual polish and return a recommendation with alternatives considered. - File-Aware Context: Run from the repo root so claude can read referenced files by relative path when forming its judgment. - macOS Keychain Authentication: On macOS, automatically uses the long-lived Claude Code OAuth token stored in Keychain when present, falling back to the machine's normal Claude authentication otherwise. - Use Case: You are unsure whether a settings panel label should read "Sync Preferences" or "Synchronization Options". The Skill prompts Claude with the question and relevant file paths, receives a recommendation with reasoning, and applies it. ## Quick Start Ask the taste-review skill to decide between two naming options for a function, referencing the relevant source files in the repository.

Frequently Asked Questions about taste-review

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

FAQPage Schema
How do I get an AI opinion on naming or phrasing decisions?

Run the claude CLI with a prompt stating your question plainly, listing relevant file paths, and asking for a recommendation with alternatives considered. The response gives a reasoned judgment call you can apply directly.

How do I use the claude CLI non-interactively in a script?

Use claude -p followed by your prompt string to run Claude in print mode without an interactive session. Build the prompt with a heredoc so multi-line questions and file references are passed cleanly.

Does the claude CLI work with macOS Keychain OAuth tokens?

Yes. On macOS, retrieve the token with security find-generic-password for the Claude Code skill OAuth entry, then pass it via the CLAUDE_CODE_OAUTH_TOKEN environment variable. Without it, the CLI falls back to normal authentication.

Why should I run claude from the repository root?

Running from the repo root lets claude resolve relative file paths mentioned in the prompt, so it can read the actual files under discussion. Running elsewhere breaks relative path resolution and weakens the recommendation.

When should I not delegate a decision to an AI judgment call?

Avoid delegating decisions with objective correctness criteria, security implications, or measurable performance requirements. Taste-driven delegation suits subjective choices like wording, naming, and visual polish where multiple answers are defensible.