daily-account-prioritizer-v2

Daily Account Prioritizer v2

This skill produces a ranked, scored list of accounts for an SDR or AE to act on

today. It uses the reo_get_daily_focus_list composite endpoint — a single API

call that handles all data fetching, filtering, scoring, and enrichment server-side.

Config

• TERRITORY_DEFAULT: ask user
• TOP_N: 20
• RECENCY_DAYS: 7
• EXCLUDED_CRM_STAGES: Customer, Closed Lost, Churned, Disqualified
• PROSPECT_SENIORITY: Founder, C-Level, VP, Director, Head

Claude's Job

1. Confirm the user's territory
2. Call reo_get_daily_focus_list once
3. Write "Why it's here" and "Recommended action" for each account
4. Render the HTML dashboard
5. Provide a narrative summary

Scoring Overview (100 points total)

The composite endpoint computes all scores server-side. Claude does NOT need to

calculate any of these — they arrive pre-computed in the response. This reference

is for understanding the dashboard display:

| Component | Weight | What it measures |

| -------------- | ------ | ------------------------------------------------------------------- |

| ICP fit | 20 | Customer fit tier (STRONG=20, MODERATE=10) |

| Funnel fit | 25 | Developer stage (DEPLOY=25, BUILD=17, etc.) |

| Activity score | 25 | Developer activity score × 0.25, capped at 25 |

| Recency score | 30 | Activity delta (15) + Mgr+ identified (10) + Δ deanon devs (5, n/a) |

Max attainable: 95 (Δ deanon dev scoring is unavailable).

Step 0 — Confirm Territory

Ask once: "What's your territory? (e.g. US + Canada, EMEA, or 'global')"

Wait for the answer unless it's obvious from context or a prior conversation.

Map the answer to a country_list array:

• "US + Canada" → ["United States", "Canada"]
• "EMEA" → ["United Kingdom", "Germany", "France", "Netherlands", ...]
• "global" → omit the parameter entirely

Step 1 — Fetch the Focus List (single call)

Call reo_get_daily_focus_list with:

reo_get_daily_focus_list(

  country_list = ["United States", "Canada"],   // from Step 0

  top_n = 20,

  recency_days = 7,

  customer_fit_score = ["STRONG", "MODERATE"],

  excluded_crm_stages = ["Customer", "Closed Lost", "Churned", "Disqualified", "CUSTOMER"]

)

This single call returns everything:

| Response Field | What It Contains |

| --------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------- |

| meta | Timestamp, territory, filters applied, pool stats |

| meta.pool_stats | total_accounts_fetched, after_hard_filters, stage_a_shortlist, final_top_n |

| accounts[].scores | Pre-computed: icp_fit, funnel_fit, activity_score, stage_a, delta_activity, delta_activity_pts, mgr_plus, mgr_plus_name, mgr_plus_pts, recency, composite |

| accounts[].top_activities | Up to 3 recent high-signal activities per account |

| accounts[].developers | Up to 3 deanonymized developers with email, title, LinkedIn |

| accounts[].prospects | Up to 3 LinkedIn prospects with tech_function_match flag |

If the response has fewer than 5 accounts, note it in the dashboard banner. The

endpoint has already applied all filters — a small result set means the territory

or recency window is genuinely narrow.

Step 2 — Write Account Intelligence

For each account, Claude adds two pieces of natural language intelligence that

the endpoint cannot generate:

"Why it's here" — one concrete sentence per account

This must be specific to actual signals, people, and dates. It is the single most

important line on each card — an SDR will decide whether to act based on this.

Good examples:

• "Wes Anderson copied from the documentation page this week while the CEO's activity score hit 100 — active evaluation."
• "CMO Gaurav Deshpande has a perfect activity score and Shaneeza Mohamed is building automations — multiple teams engaged."
• "Founder Ting Wang is personally active on the platform with 100 activity score — founder-led evaluation."

Bad examples (never produce these):

• "This account has high activity." ← too generic
• "Strong ICP fit and good engagement." ← says nothing actionable

"Recommended action" — one prescriptive line per account

Base this on the combination of funnel stage, scores, and people identified:

| Signal Pattern | Action Template |

| ----------------------------------------- | -------------------------------------------- |

| DEPLOY + high recency | "Call today — expansion signal" |

| DEPLOY/BUILD + Mgr+ identified | "Book a meeting with [mgr_plus_name]" |

| High activity + multiple developers | "Multi-thread: reach [dev1] and [prospect1]" |

| DISCOVER + strong ICP + rising activity | "Send targeted content + monitor this week" |

| Strong ICP + flat activity | "Warm outbound to [prospect name]" |

| High activity + MODERATE ICP | "Qualify before investing time" |

| prospects with tech_function_match = true | Prioritize those contacts in the action |

Step 3 — Render the HTML Dashboard

Generate a single self-contained dark-mode HTML dashboard file. Save to outputs

folder and share the link.

Header

• Title: "🎯 Today's Focus List — [Date]"
• Subheader: territory · eligible pool size · shortlist size · data freshness
• Legend: ICP 20 · Funnel 25 · Activity 25 · Recency 30
• Banner: "Δ deanon dev scoring (3E) unavailable — max attainable score: 95" (from meta.scoring_notes)

One Card Per Account (ranked 1–N)

Each card contains:

1. Header row — rank, company name, domain, composite score with a horizontal bar. Bar colours: green ≥ 70 · amber 45–69 · red < 45.
2. Score breakdown — four pills: "ICP 20/20 · Funnel 25/25 · Activity 18/25 · Recency 22/30" Compact recency line: "Δ activity +15 · Mgr+ ✓ [name]" or "Mgr+ ✗"
3. "Why it's here" — the sentence from Step 2.
4. Top activities — from accounts[].top_activities. Each row: Fetch the accounts activity summary of last 7 days. If empty (upstream API unavailable): show "Activity detail temporarily unavailable"
5. Known developers — from accounts[].developers. Each row: name · title · signal_strength · LinkedIn ↗ If title is empty, show email instead. If no developers: "— not yet identified"
6. More prospects — from accounts[].prospects. Each row: name · title · seniority · LinkedIn ↗ If tech_function_match is true, show a small badge: "✓ tech match" If no prospects: "— none found at this seniority level"
7. Recommended action — the prescriptive line from Step 2.

Styling

• Background #0f0f12 · card #1a1a22 · border #27272e
• Accent #6366f1 · text #e5e7eb · muted #9ca3af
• Score bar: green #22c55e · amber #f59e0b · red #ef4444
• Tech match badge: #14532d bg, #4ade80 text
• Readable sans-serif, generous whitespace. No emoji soup.
• Max-width 900px centered. Cards with border-radius: 12px.

Step 4 — Narrative Summary

3–5 sentences in chat alongside the dashboard link:

• Dominant pattern across the top accounts (e.g., "heavy on DEPLOY-stage accounts")
• Single most urgent action and why
• Notable outlier (e.g., a cold account that suddenly appeared, a founder personally evaluating)
• Quick note on pool health: how many survived filters vs total

Edge Cases

• Fewer than 5 accounts returned: show a banner noting the pool is thin. Suggest the user try widening territory or relaxing the recency window.
• top_activities empty for some accounts*: the upstream activities API may be temporarily unavailable for some accounts. Show "Activity detail temporarily unavailable" instead of leaving the section blank.
• All tech_function_match = false: common for small companies where leadership is tagged "Not sure". Don't flag this as a problem — just omit the badge and show prospects normally.
• org_name blank*: show domain only, labelled "(unnamed)".
• mgr_plus_name references "QA Lead" or similar non-manager title*: the endpoint's keyword matching is broad. Don't second-guess it in the dashboard — show the name as-is.

Technical Notes

Response field reference — exact paths in the reo_get_daily_focus_list response:

• meta.generated_at — ISO timestamp
• meta.territory — array of countries or ["global"]
• meta.pool_stats.total_accounts_fetched — raw pool size
• meta.pool_stats.after_hard_filters — after CRM + recency filters
• meta.pool_stats.final_top_n — number of accounts in response
• meta.scoring_notes[] — array of strings, display as banners

Per account:

• rank, org_id, domain, org_name, industry, location
• estimated_employee_range, developer_stage, customer_fit
• deanonymised_dev_count_range
• scores.icp_fit (max 20), scores.funnel_fit (max 25)
• scores.activity_score (max 25), scores.stage_a (max 70)
• scores.delta_activity (raw delta), scores.delta_activity_pts (max 15)
• scores.mgr_plus (bool), scores.mgr_plus_name (string)
• scores.mgr_plus_pts (0 or 10), scores.recency (max 25 effective)
• scores.composite (max 95)
• top_activities[] — {actor_name, activity_type, page, activity_source, timestamp}
• developers[] — {name, email, title, linkedin, signal_strength, activity_score}
• prospects[] — {name, title, seniority, tech_function, location, linkedin, email, tech_function_match}

Activity signal hierarchy (for narrating "Why it's here"):

FORM_CAPTURE > DEMO_LOGIN > COPY_TEXT/COPY_COMMAND > PAGE_VISIT > Identity

Relative time display: convert timestamp to relative format for the dashboard:

• Today → "today", yesterday → "1d ago", etc. Use activity_source date, not render time.