# Review Monitor
Gather competitor reviews from the open web and summarize sentiment and themes into a file for the team to act on.

## When to use
- A stakeholder asks for a fresh read on how a competitor's product is being received.
- You need a periodic pulse check on review-site sentiment (G2, Capterra, Trustpilot, app stores, Reddit threads).
- A competitor just launched a feature or had an incident and you want to see if reviews reflect it.
- Someone requests a written brief comparing sentiment across two or more competitors.

## Tools
- `browser-search`: find review pages, forum threads, and aggregator listings for the target competitor.
- `browser-navigate`: open the specific review pages, app store listings, or forum threads surfaced by search.
- `browser-extract`: pull the review text, star ratings, dates, and reviewer context off each page.
- `write-file`: save the compiled sentiment summary and theme breakdown as a durable artifact.

## Playbook
1. Confirm the competitor name, product, and any specific time window or feature focus from the request.
2. Run `browser-search` for the competitor name plus "reviews" across common sources: G2, Capterra, Trustpilot, app stores, and Reddit.
3. For each promising result, use `browser-navigate` to open the page.
4. Use `browser-extract` to capture review text, star ratings, dates, and any reviewer role/company context visible on the page.
5. Repeat steps 2-4 across at least three distinct sources so the sample is not skewed by one platform's audience.
6. Group the extracted reviews into recurring themes (for example: pricing complaints, support quality, onboarding friction, standout features) and note the rough sentiment split (positive/neutral/negative) per theme.
7. Draft a concise summary: overall sentiment trend, top 3-5 themes with representative quotes, and any notable spikes tied to recent product changes.
8. Save the summary with `write-file`, naming the file clearly with the competitor and date range covered.
9. If the request implies ongoing tracking, note in the summary what should be re-checked next cycle and flag it via `send-message` to a human for scheduling, since this skill only runs on demand.

## Failure modes
- If a review site blocks navigation or extraction returns empty content, fall back to `browser-search` with narrower queries or alternate sources rather than reporting a false gap.
- If a page structure has changed and `browser-extract` pulls garbled or irrelevant content, verify by re-navigating and adjusting what you extract before including it in the summary.
- If searches return no results for a competitor, double-check the spelling and try known alias names or parent company names before concluding there is no review presence.
- If you hit repeated rate limiting or blocking from a source, back off, continue with other sources, and note the gap explicitly in the final file rather than fabricating data.

## Done when
- A file exists (via `write-file`) containing the sentiment summary, theme breakdown, and source list with dates.
- The summary draws from at least three distinct review sources to avoid single-platform bias.
- Any follow-up tracking need has been flagged rather than silently dropped.