Brandwatch crisis early-warning agent for Slack

Build me an agent-based workflow that acts as a Brandwatch crisis early-warning system for our brand and posts structured alerts into a Slack channel when something starts trending against us.

Trigger: cron, every hour during business hours in our timezone (for example Monday to Friday, 8am to 7pm). The agent should stay silent on weekends and outside business hours.

On every run, the agent should:

1) Use Brandwatch List Projects to find the project the comms team uses for brand monitoring, then Brandwatch List Queries to find the brand-monitoring query inside that project. Resolve the project and query by name (the names should be configurable inputs to the workflow, for example project name 'Brand Monitoring' and query name 'Acme Brand'). Cache these IDs across runs if possible so we are not re-resolving them every hour.

2) Call Brandwatch Get Chart Data twice against that query: once with the days dimension to pull daily mention volume for the last 8 days, and once with the days x sentiment dimensions to pull daily counts split by sentiment (positive, neutral, negative) for the same 8 days. Use the most recent full day as 'today' and the 7 days before that as the baseline.

3) Compute two signals: (a) a volume spike when today's mention count is more than 2x the 7-day baseline average, and (b) a negative-sentiment surge when today's share of negative mentions is materially higher than the baseline share (for example >1.5x the baseline negative share, or negative share >30% with a meaningful absolute volume). Either signal counts as a crisis trigger. Make these thresholds easy to edit at the top of the instructions.

4) If neither signal fires, do nothing and exit silently. Do not post a 'nothing to report' message.

5) If a signal fires, call Brandwatch Retrieve Mentions for the same query, scoped to the last 24 hours and filtered to negative sentiment, sorted by reach descending. Take the top 3 mentions. For each, capture the author handle, the audience reach, the URL, and a short excerpt of the post text (trim to roughly 240 characters).

6) Compose a Slack alert and post it to a designated channel (default #brand-alerts, but make the channel name a configurable input) using Slack Bot Send a Message. The message should include: a clear headline like ':rotating_light: Brand crisis signal: Acme Brand' with the offending signal type, a one-line summary of why it spiked (for example 'Mentions are 3.4x the 7-day average, with 58% negative sentiment vs 22% baseline'), the top 3 example mentions as bullets with author, reach, the trimmed excerpt, and the URL, and a 'Recommended next step' line such as 'Have comms review the top driver post and decide on a public response within 30 minutes.' Format with Slack mrkdwn so links are clickable and key numbers are bold.

7) Keep the message tight and skimmable. Cap example mentions at 3 even if more are negative. Avoid posting raw API data or IDs. The reader is the head of comms, not an analyst.

Inputs the workflow should expose: Brandwatch project name, Brandwatch query name, Slack channel, business-hours window and timezone, volume-spike multiplier (default 2.0), negative-sentiment multiplier (default 1.5), absolute negative-share floor (default 30%), and an optional Slack user or group to @-mention for severe alerts (for example when volume is >5x baseline).

Be defensive: if Brandwatch returns no data for the baseline window, skip the spike check for that run rather than alerting on division-by-zero. If the query or project name cannot be resolved, post a one-time setup error to the same Slack channel and stop.