Menu
Ask Claude Code to make a taste-driven call on something ambiguous — UI polish, prose phrasing, naming, formatting. Use when you'd otherwise guess.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Add a user-facing entry to CHANGELOG.md for work that just landed. Use when wrapping up a feature/fix, or when asked to record what changed. Called standalone or from the `done` skill.
Implement a fix or feature from a GitHub, Jira, Linear, or other issue-tracker issue by fetching issue context, inspecting the current codebase, making code changes, validating them, opening a GitHub pull request, and reporting progress back to the original issue.
Prepare Hubble desktop releases using tag-triggered GitHub Releases. Use when cutting, automating, or explaining app releases, version bumps, tags, GitHub Actions publishing, Electron artifacts, or updater release management.
Review a pull request or local branch diff for correctness, security, lifecycle, error handling, tests, and meaningful performance risks. Use when asked to review PR changes, a branch, a commit range, or the current working tree and return a structured review.
Request an automated review of the current pull request, address valid feedback, and repeat until approval. Use after opening a PR and before handing it to a human reviewer.
Open a pull request for the current branch. Use whenever asked to make/create/open a PR.
| name | taste-review |
| description | Ask Claude Code to make a taste-driven call on something ambiguous — UI polish, prose phrasing, naming, formatting. Use when you'd otherwise guess. |
You hit something fuzzy and need a judgment call. Shell out to the claude CLI to get one back, then apply it.
Run from the repo root so claude can read files by relative path:
Run outside the sandbox when required, requesting reusable approval for the
claude -p prefix. On macOS, use the optional long-lived subscription OAuth
token from Keychain when present. Otherwise, fall back to that machine's normal
Claude authentication.
prompt="$(cat <<'EOF'
<your question, stated plainly>
Files to consider: <paths, if any>
Weigh a few options, give your recommendation, and share others as alternatives considered.
Length is up to you — a design call may warrant several paragraphs;
a naming call may not. Match the depth to the decision.
EOF
)"
if [[ "$(uname)" == "Darwin" ]] &&
oauth_token="$(
security find-generic-password \
-a "$USER" \
-s "Claude Code skill OAuth" \
-w 2>/dev/null
)"; then
CLAUDE_CODE_OAUTH_TOKEN="$oauth_token" claude -p "$prompt"
else
claude -p "$prompt"
fi