agentsclimarketplace

Review responder

Skill buildwithreef/claude-skills/skills/review-responder

12 free Claude Skills for small business — invoice chasing, review replies, local SEO audits, landing page teardowns, and more. Real SOPs, not prompts.

Install
npx -y skills add buildwithreef/claude-skills --skill review-responder

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Writes owner responses to Google, Yelp, and Facebook reviews — positive, negative, or fake — in an authentic local-business voice. Use when the user wants to reply to a customer review, handle a bad review, deal with a fake or wrong-business review, or says things like "how should I respond to this review", "we got a 1-star", or "write a reply to this Google review".

SKILL.md

4.5 KB, 961 tokens by cl100k_base, as published. Nobody here has run it

Review Responder

Write review replies that sound like the owner typed them between jobs — because to future customers, the reply matters more than the review.

Step 1 — Read the review, then classify it

Ask for the review text, star rating, platform, and what actually happened (the owner's side). Then classify:

TypeSignalsGoal of the reply
Positive (4–5★)Real detail, named staffReinforce the detail, invite back
Negative, legitimateSpecific complaint the owner recognizesOwn it, fix it, move it offline
Negative, vague1★, little or no textShow readers you care, ask for specifics
Fake / wrong businessNo record of the customer, wrong detailsCalmly flag it for readers, report it
Real customer, false claimsRecognizable job, distorted storyCorrect facts once, no fighting

Never draft until you know whether the reviewer was actually a customer and what the owner's side is. If the user just pastes a review, ask those two things first.

Step 2 — Apply the voice rules

  • Sign with the owner's first name. Replies from "The Management Team" convince no one.
  • Echo one specific detail from the review. It proves a human read it.
  • Positive replies: 1–3 sentences. Gushing at a 5-star review reads as desperate.
  • Never open two replies in a row the same way. Vary the opener or skip the thank-you.
  • No coupons, discounts, or refunds in public replies — it teaches people that complaining publicly pays.
  • Write for the next 100 readers, not the one reviewer. They are the real audience.
  • Medical, dental, legal, and financial businesses: never confirm the person was a patient or client (HIPAA / confidentiality). Reply in generalities: "Privacy rules mean I can't discuss who we do or don't treat here, but anyone with a concern can reach me directly at [number]."

Step 3 — Structure by type

Negative, legitimate (the one that matters most):

  1. Name the specific issue. Banned: "we regret any inconvenience" and every phrase shaped like it.
  2. Own your share plainly: "We were late. That's on us."
  3. Say what changed or what you'll do about it.
  4. Move it offline with a direct line: "Call me at [number] and I'll make it right."
  5. Keep it 3–5 sentences. Longer reads as defensive.

Fake or wrong business: reply for the audience, not the reviewer — "I've checked our records and can't find any customer matching this review. If you have us mixed up with another business, no hard feelings — but if you believe this is us, call me directly at [number]." Then report it (Google: flag as "not a real experience"; Yelp: Report Review) and tell the owner removal takes days to weeks — the reply is what readers see in the meantime.

Real customer, false claims: correct the record once with facts and dates, no adjectives — "Our records show we quoted $450 on March 3 and the final invoice was $450." — then the offline invitation. Never trade blows in a thread.

Worked examples across industries (salon, restaurant, HVAC, dentist, auto shop) are in references/playbook.md. Load it before drafting.

Step 4 — Deliver

**Type:** Negative, legitimate — 2★ on Google
**Reply:**

[reply text]

**Why this works:** [1–2 bullets]
**Also do:** [report/flag step, or the operational fix the review exposed]

If the review exposes a real operational problem (chronic lateness, a rude employee, quality slip), say so in "Also do." The reply is damage control; the fix is the actual work.

What not to do

  • Don't reply while the owner is still angry. If the situation is hot, draft it and recommend posting after 24 hours.
  • Don't argue point-by-point. One factual correction maximum.
  • Don't stuff keywords ("Best plumber in Des Moines thanks you!") — reads as spam to humans, does nothing for rankings.
  • Don't ignore star-only reviews with no text — a one-line reply still shows future readers the owner is present.
  • Don't use the reviewer's full name in medical or sensitive service contexts.

Built by Reef — AI systems for small business → https://buildwithreef.com

Gives 0 of the 12 instructions most review quality skills give in 961 tokens

Counted across 1,048 of the 1,783 authors here whose files we hold, read 2026-08-06

  • ask questions one at a timein 82 of 1048, across 54 files
  • provide a recommended answer for each questionin 73 of 1048, across 45 files
  • explore the codebase instead of asking answerable questionsin 66 of 1048, across 37 files
  • resolve dependencies between decisions one-by-onein 42 of 1048, across 15 files
  • interview the user relentlessly about the planin 39 of 1048, across 12 files
  • order findings by severityin 29 of 1048
  • resolve each branch of the decision treein 28 of 1048, across 5 files
  • run a grilling sessionin 26 of 1048, across 5 files
  • update CONTEXT.md immediately when a term is resolvedin 26 of 1048, across 9 files
  • propose precise canonical terms for vague languagein 25 of 1048, across 6 files
  • create documentation files lazilyin 24 of 1048, across 5 files
  • use the domain-modeling skillin 22 of 1048, across 3 files

Said here and by no other author read

  • Ask for the review text and owner's side
  • Classify the review type before drafting
  • Load the reference playbook before drafting
  • Sign with the owner's first name
  • Echo one specific detail from the review
  • Keep positive replies between one and three sentences

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.