cosmicstack-labs/review-responder
>- Monitor customer reviews across platforms and generate thoughtful, brand-aligned responses. Flags urgent negative reviews for priority attention, drafts responses that turn detractors into promoters, and tracks review sentiment trends over time.
npx skills add https://github.com/cosmicstack-labs/mercury-agent-skills --skill review-responder
Online reviews are the digital word-of-mouth that makes or breaks a shop or restaurant. A single 1-star review seen by 100 potential customers can cost thousands in lost revenue — but a thoughtful, timely response can neutralize the damage and even turn a critic into a loyalist. This skill treats every review as a reputation management opportunity.
Ask the user which platforms they use. Common sources:
For each platform, collect:
Classify each review:
| Category | Rating | Sentiment | Response Priority |
|----------|--------|-----------|-------------------|
| 🔴 Crisis | 1 star | Angry, accuses of health/safety issue, public figure | Immediate (< 1 hr) |
| 🟠 Critical | 1-2 stars | Significant complaint, specific issue, or repeat detractor | Same day |
| 🟡 Mixed | 3 stars | Balanced feedback — some praise, some criticism | Within 24 hrs |
| 🟢 Positive | 4-5 stars | Happy customer, specific praise | Within 48 hrs |
| ⚪ Neutral | Any | Factual, no emotion (e.g., "They exist") | Low priority |
If a crisis review is detected:
> 🚨 ALERT: 1-star review on Google mentions "found a foreign object in food." This is a health/safety escalation. Recommend immediate owner response and internal investigation before replying publicly.
🔴 Crisis / 🟠 Critical (1-2 stars)
Structure: Acknowledge → Apologize → Explain (briefly) → Make it right → Move offline
Dear [Name],
Thank you for bringing this to our attention. I'm truly sorry your experience
at [Business Name] didn't meet expectations — [specific reference to their
complaint, e.g., "the wait time on Friday night"] is not the standard we aim for.
We've already [specific action taken, e.g., "reviewed our Friday staffing
plan to ensure faster seating"]. I'd love the chance to make this right
personally. Please email me at [manager email] so I can hear more and
arrange a visit on us.
Best,
[Manager Name]
🟡 Mixed (3 stars)
Structure: Thank → Acknowledge specifics → Address what they flagged → Invite another visit
Hi [Name],
Thanks for the balanced feedback. We're glad you enjoyed [specific praise,
e.g., "the burger and fries"] — and we hear you on [specific criticism,
e.g., "the slow service during the dinner rush"]. We've taken note and
will work on [specific improvement].
We'd love to have you back to show you our best. Next time, mention this
message and we'll [small gesture, e.g., "start you with a drink on us"].
Cheers,
[Manager Name]
🟢 Positive (4-5 stars)
Structure: Thank → Reference specifics → Personalize → Invite return
Hi [Name],
Thank you so much for the kind words! We're thrilled you loved [specific
item or experience, e.g., "the truffle fries and the atmosphere"]. [Staff
member name] was especially happy to hear your shout-out.
Looking forward to serving you again soon!
Best,
[Manager Name]
After processing all reviews, generate a trend summary:
📊 REVIEW SUMMARY — March 2025
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Total Reviews: 23
Average Rating: 4.2 ⭐
Platforms: Google (14), Yelp (6), TripAdvisor (3)
SENTIMENT BREAKDOWN
Positive (4-5): 16 (70%)
Mixed (3): 5 (22%)
Critical (1-2): 2 (8%) ← Up from 4% last month
COMMON THEMES
👍 Praised: "friendly staff" (8 mentions), "great coffee" (6)
👎 Flagged: "wait time" (5 mentions), "parking" (4)
TREND: Rating trending down slightly. Wait time complaints increasing.
RECOMMENDATION: Consider adding one more server during peak hours (Fri-Sat 7-9pm).
For reviews that reveal operational issues:
> Internal note: "Cold food mentioned in 3 reviews this week. Check: is the pass heater working? Are expo staff calling ready plates promptly? Address with kitchen team."
> "I can check your reviews daily and flag anything urgent. Would you like me to set that up? I'll prioritize critical reviews and draft responses for your approval."
| Phrase | What Happens |
|--------|-------------|
| "Check my reviews" | Fetches recent reviews from connected platforms |
| "Draft a response to [name/review]" | Generates a personalized response draft |
| "Any new bad reviews?" | Shows critical reviews requiring immediate attention |
| "How are my reviews this month?" | Generates sentiment trend report |
| "Reply to all reviews from yesterday" | Batch drafts responses for unaddressed reviews |
| "What are customers complaining about?" | Extracts common themes from recent negative reviews |
| "Share our best review" | Formats a glowing review for social media sharing |
*"Every review is a gift — even the painful ones. They tell you exactly where to improve."*
Take cosmicstack-labs/review-responder from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.