cc-daily-brief

name	cc-daily-brief
description	Use when the user asks for a contact-centre daily brief, morning brief, daily standup data, or 'what happened yesterday' — e.g. 'daily brief', 'morning brief for yesterday', 'what happened in the CC yesterday', 'how did we go yesterday', 'daily CC report'. Produces a one-page HTML brief covering yesterday's KPIs vs the rolling 7-day median, top flagged agents, worst routes (queues with SL drops), repeat-caller hotlist, and adherence flags. Reads tenant-specific knobs (flag thresholds, comparison window, output filename) from ~/.config/genesys-mcp/tenant.yaml. Requires the genesys MCP to be connected; run the genesys-tenant-setup skill first if the tenant config doesn't exist yet.
metadata	{"version":"1.0.0"}

Contact-Centre Daily Brief

You are producing a daily contact-centre brief — a one-page HTML document a supervisor reads at start-of-day to know yesterday's headline numbers, where issues are, and what to act on today. Tighter scope than cc-monthly-report (which is leadership-facing and covers a month).

Tenant-agnostic. All thresholds (sentiment dip, AHT excess, SL drop) come from cfg.daily_brief.flag_thresholds. Comparison window is cfg.daily_brief.comparison_window_days (default 7).

Before starting

1. Confirm genesys MCP is connected — mcp__genesys__* tools available.

2. Confirm tenant.yaml exists at the resolved config path; if missing, stop and tell the user to run genesys-tenant-setup first.

3. Read the tenant config knobs you'll need:

   cd ~/code/genesys-mcp && .venv/bin/python -c "
   from genesys_mcp.tenant import load_config
   import json
   cfg = load_config()
   print(json.dumps({
       'tenant_name': cfg.tenant.name,
       'timezone': cfg.tenant.timezone,
       'brands': cfg.brands.names,
       'mu_ids': cfg.management_units.ids,
       'bu_id': cfg.business_unit.id,
       'specialist_roles': cfg.specialist_roles,
       'targets': {'voice_aht_s': cfg.targets.voice_aht_s, 'message_aht_s': cfg.targets.message_aht_s},
       'daily_brief': {
           'comparison_window_days': cfg.daily_brief.comparison_window_days,
           'thresholds': cfg.daily_brief.flag_thresholds.model_dump(),
       },
   }, indent=2))
   "
   

4. Confirm the target day. Default is "yesterday" in cfg.tenant.timezone. Accept: "yesterday", "today", "last Monday", "2026-05-20". If the user just says "daily brief" with no day, default to yesterday.

Procedure

Step 1 — Resolve the target day's interval + the comparison window's interval

Convert the target day to [start_local 00:00, end_local 24:00) in cfg.tenant.timezone, then to UTC for the Genesys API. Use Python's zoneinfo:

from datetime import datetime, timedelta, timezone
from zoneinfo import ZoneInfo
from genesys_mcp.tenant import load_config

cfg = load_config()
tz = ZoneInfo(cfg.tenant.timezone)

# target day
day_local_start = datetime(2026, 5, 20, 0, 0, tzinfo=tz)  # parsed from user input
day_local_end   = day_local_start + timedelta(days=1)
day_interval = f"{day_local_start.astimezone(timezone.utc).isoformat().replace('+00:00','.000Z')}/{day_local_end.astimezone(timezone.utc).isoformat().replace('+00:00','.000Z')}"

# rolling-N-day comparison interval (excluding the target day itself)
window = cfg.daily_brief.comparison_window_days
window_end_local = day_local_start
window_start_local = window_end_local - timedelta(days=window)
window_interval = f"{window_start_local.astimezone(timezone.utc).isoformat().replace('+00:00','.000Z')}/{window_end_local.astimezone(timezone.utc).isoformat().replace('+00:00','.000Z')}"

Save both intervals plus a date_slug (YYYY-MM-DD) for the output filename.

Step 2 — Pull data in parallel

Make all of these calls in parallel — they're independent. Save each result's JSON payload to /tmp/cc-daily-brief-{date-slug}/:

File	Tool call	Notes
queue_perf_day.json	mcp__genesys__queue_performance(interval=<day>, queue_ids=[])	yesterday's voice + message SL by queue
queue_perf_window.json	mcp__genesys__queue_performance(interval=<window>, queue_ids=[])	rolling-N-day baseline
agent_perf_day.json	mcp__genesys__agent_performance(interval=<day>, user_ids=[])	yesterday's per-agent AHT/answered
repeat_callers.json	mcp__genesys__repeat_caller_deep_dive(queue_ids=[], interval=<day>, top_n=10)	unresolved-from-yesterday hotlist
break_overrun.json	mcp__genesys__break_overrun_report(interval=<day>, user_ids=[])	break/pre-break overruns
nps.json (v1.11, optional)	mcp__genesys__search_conversations_by_attribute(attribute_key=cfg.survey.nps_attribute_key, interval=<day>)	Only call when cfg.survey.nps_attribute_key is set in tenant.yaml. Powers the NPS card. Omit the file entirely (don't write null) when the tenant hasn't opted in — the build script gates on file presence.
wrap_up_distribution.json (v1.11)	mcp__genesys__wrap_up_code_distribution(interval=<day>, include_trend=True, top_n=5)	Top wrap-up codes + largest mover (vs immediately prior day). Section auto-omits if zero conversations.

The skill's pattern matches cc-monthly-report Step 3: parallel tool calls, save raw JSON, then a single Python build script does aggregation + HTML rendering.

Soft-fail handling (v1.12.1): If wrap_up_code_distribution (or any other v1.11 sidecar tool) returns a canonical soft-fail envelope (status >= 400), save it to the file as-is. The build script renders a visible "⚠️ data not retrieved" callout automatically — do NOT write narrative paragraphs explaining the gap in chat.

Step 3 — Run the build script

Resolve the output path from the tenant config and run build_report.py:

OUTPUT_PATH=$(cd ~/code/genesys-mcp && .venv/bin/python -c "
from genesys_mcp.tenant import load_config
print(load_config().daily_brief_output_path('{date-slug}'))
")

python ~/code/genesys-mcp/skills/cc-daily-brief/build_report.py \
  --target-date "{date-slug}" \
  --day-interval "{day-interval}" \
  --window-interval "{window-interval}" \
  --data-dir /tmp/cc-daily-brief-{date-slug} \
  --output "$OUTPUT_PATH"

The script loads tenant.yaml, reads each JSON file from the data directory, computes yesterday-vs-rolling-median deltas per queue, surfaces flagged agents per flag_thresholds, picks worst routes by SL drop, builds the repeat-caller hotlist, and writes the HTML.

Step 4 — Synthesise narrative sections (v0.9+)

Open the freshly-generated HTML and skim the data sections — headline KPIs, flagged routes, flagged agents, callbacks, adherence. Use those numbers (not your prior assumptions) to draft 2 short narrative sections that go at the top of the brief for the supervisor's morning glance:

## Headline

One paragraph (≤ 80 words). What's the headline of yesterday? Use the **rolling median** as the comparison anchor — *"Voice SL 65% (rolling 78%), driven by a Tuesday 10am drop where 3 of 4 eligible specialists were on extended interactions"*. Name the **one or two things that explain most of the variance**. Don't list every flag — that's what the data sections below are for.

## Today's priorities

Top 3 actions for **today**, not yesterday. Sorted by impact-per-effort. Each one bullet:

- **Action — owner / effort** — *brief evidence from the data*
- **Action — owner / effort** — *brief evidence*
- **Action — owner / effort** — *brief evidence*

Save to /tmp/cc-daily-brief-{date-slug}/narrative.md, then re-run the build script with the --with-narrative flag pointing at it:

python ~/code/genesys-mcp/skills/cc-daily-brief/build_report.py \
  --target-date "{date-slug}" \
  --day-interval "{day-interval}" \
  --window-interval "{window-interval}" \
  --data-dir /tmp/cc-daily-brief-{date-slug} \
  --output "$OUTPUT_PATH" \
  --with-narrative /tmp/cc-daily-brief-{date-slug}/narrative.md

The build script parses by ## Heading, runs each section's body through a minimal markdown subset (paragraphs, bold, italic, code, [links], - bullets), and slots the combined narrative as a single "Daily summary" section at the top of the brief, above the data sections.

If the user explicitly says "skip the narrative" or yesterday's data is genuinely unremarkable (everything within the rolling median bands), omit the --with-narrative flag and ship the data-only brief — that's the v0.7-era behaviour and it's still a valid output.

Step 5 — Confirm + brief

After the script succeeds, post a short summary in chat (don't paste HTML):

• Output path
• Yesterday's headline: voice SL %, total interactions, vs the rolling median
• The top-3 flagged agents (one-liner each, with the flag reason)
• 1-2 worst routes if any (queue + SL drop)
• Any unresolved repeaters that should get callbacks today

Keep it tight — one paragraph plus a 3-5-item bullet list. The HTML is the deliverable; the chat summary is the "what should I do first today" hook.

What the HTML contains

Single-page, ~700px wide, designed to fit a laptop screen or a Slack share without scrolling much. Sections:

1. Headline KPIs — voice SL, message SL, total interactions, vs the rolling-N-day median (colour-coded .vs-target pills)
2. Worst routes — top 3-5 queues by SL drop vs their rolling median, with current eligible-agent counts where helpful
3. Flagged agents — top 3-5 agents by composite flag score (AHT excess + sentiment dip + adherence)
4. Repeat-caller hotlist — unresolved-from-yesterday repeaters who should get a callback today
5. Adherence flags — agents with break/pre-break overruns > 30 min

Same visual idiom as cc-monthly-report (CSS reused) — colour-coded vs-target pills, KPI cards, no JavaScript. Print-friendly.

When NOT to use this skill

• If the user wants a multi-day or week-level view, use cc-monthly-report or wait for explicit weekly variants
• If the user wants a deep dive on one agent, use cc-coaching-prep
• If the user wants the data raw (e.g. for Excel pivot), call the underlying MCP tools directly

Configurable behaviour

Knob	Source	Notes
Sentiment dip threshold	cfg.daily_brief.flag_thresholds.sentiment_dip	default 0.4
AHT-excess threshold	cfg.daily_brief.flag_thresholds.aht_excess_pct	default 15% over voice AHT target
SL-drop threshold	cfg.daily_brief.flag_thresholds.sl_drop_pp	default 10 percentage points
Comparison window	cfg.daily_brief.comparison_window_days	default 7 days
Output filename	cfg.daily_brief.output_filename_pattern	default daily-brief-{date}.html