# Mention Monitoring
Track brand mentions across platform APIs and the open web, then notify the right people when a mention is notable.

## When to use
- A scheduled sweep for brand, product, or executive name mentions is due.
- A spike in mention volume or an unusually negative mention needs triage.
- A specific keyword, hashtag, or competitor comparison must be monitored on demand.
- A prior mention thread needs a follow-up check for new replies or engagement.

## Tools
- `http-get`: pull mentions from platform APIs (social/listening tools) via the org's connected Integration.
- `browser-search`: find open-web mentions (news, forums, reviews) not covered by an API.
- `browser-extract`: pull mention text, author, timestamp, and engagement counts from a found page.
- `send-message`: notify a human or another agent when a mention is notable or needs a human decision.

## Playbook
1. Confirm the tracked terms (brand name, product names, key handles, competitors) for this run.
2. Call the platform's REST API with `http-get` (e.g. GET the mentions/search endpoint for the connected social listening or social media service). Authentication is injected by the org's Integration row for this host — never ask for, echo, or hardcode credentials. If the call returns 401/403 or no Integration row exists, stop and message a human via `send-message` asking them to create the Integration, naming the service and required scopes.
3. Run `browser-search` for the same terms across open-web sources (news sites, forums, review sites) not covered by the platform API.
4. For each promising open-web result, use `browser-extract` to pull the mention text, author/publisher, timestamp, and any visible engagement counts.
5. Normalize all mentions (platform + open web) into a single list with source, author, text snippet, link, and timestamp.
6. Score each mention for notability: high follower/authority author, negative sentiment, high engagement, or mention of a sensitive topic (pricing, outage, security, legal).
7. For each notable mention, draft a one-line summary with link and suggested urgency.
8. Send the notable mentions summary to the responsible human or agent via `send-message`, flagging any that need an urgent response.
9. If nothing notable was found, send a brief "no notable mentions" status via `send-message` so the sweep is confirmed as run, not silently skipped.

## Failure modes
- 401/403 on the platform API -> missing Integration; stop and escalate via `send-message` naming the service and scopes needed.
- Empty result set -> verify the search terms, date range, and endpoint/params before concluding there are no mentions.
- Rate limited by the platform API -> back off and retry later; note the rate limit and next safe retry time via `send-message` or memory.
- Target page structure changed and `browser-extract` fails to pull clean data -> fall back to `browser-search` for a cached snippet or alternate source.

## Done when
- A normalized mention list (platform + open web) has been produced for the run's tracked terms.
- Notable mentions have been summarized and delivered via `send-message` with links and urgency flags, or a "no notable mentions" status was sent.
- Any Integration or rate-limit blockers have been escalated to a human with the specific service and next step named.