VP Portfolio Health Report Generator

01 · Business Problem

The Problem

A VP of Engineering needs weekly portfolio health reports across 5+ Jira projects. Today this means manually exporting Jira, reviewing 50+ epics for RAG status, writing an exec narrative, formatting slides — 4+ hours of TPM time every week. Plus daily monitoring for escalations happens ad-hoc.

The Solution

A 3-agent automated pipeline:

Daily Jira Monitor — detects changes and emails delta reports

Escalation Agent — identifies Red/Critical triggers, drafts VP-ready escalation emails

Weekly Report Generator — exports Jira → Claude analysis → 5-slide PPTX deck

Total cost: ~$0.05/run

02 · Solution Description

Why Not Manual Reports?

4+ hours of TPM time every week

RAG classification is subjective and inconsistent

Escalations caught too late — after the meeting, not before

No audit trail of what changed between reports

Why 3-Agent Pipeline?

Daily monitoring catches issues within 24 hours

Escalation agent drafts VP emails with specific triggers

Claude generates consistent RAG analysis + exec narrative

Full audit trail via JSON + snapshots

03 · Technical Architecture

Model

Claude
claude-sonnet-4-6
Anthropic SDK
max_tokens=16000

Data

Jira REST API (KAN project)
CSV export
Daily JSON snapshots

Agent 1

jira_monitor.py
Daily delta detection
Email report

Agent 2

escalation_agent.py
Red/Critical triggers
VP email draft

Agent 3

run_weekly_report.py
Jira CSV → Claude → 5-slide PPTX

Output

PPTX deck
JSON audit trail
Escalation emails

04 · System Architecture 3-agent Jira → PPTX pipeline

Daily Pipeline (Agents 1 & 2)

📊

Jira REST API

KAN project · daily pull

Jira Monitor · jira_monitor.py

Compares today's snapshot vs yesterday → detects new/changed/resolved issues → emails daily delta report

Escalation Agent

escalation_agent.py

Snapshot + meeting notes → Red/Critical triggers → VP escalation email draft

📝

Meeting Notes

meeting_notes.txt

Auto-updated from live Jira data via Claude — 2 items per section (risks, wins, decisions). Fed into both escalation agent & weekly report as context.

📧

Delta Email

Daily change summary

New issues · status changes · priority shifts

Weekly Pipeline (Agent 3)

📄

Jira CSV Export

5+ projects · 50+ epics

📝

meeting_notes.txt

Auto-updated from Jira · risks · wins

Weekly Report Generator · run_weekly_report.py

Jira CSV → Claude API (RAG analysis + exec narrative) → pptxgenjs → 5-slide PPTX

claude-sonnet-4-6

5-SLIDE PPTX

VP_Portfolio_YYYY-MM-DD.pptx
Cover · Snapshot · Decisions · Wins · Narrative

JSON AUDIT

VP_Analysis_YYYY-MM-DD.json
Full Claude analysis + RAG data

ESCALATION

escalation_draft.json
VP email draft if triggers fire

meeting_notes.txt — Format & Auto-Update

📄 File Format

## Weekly TPM Sync — YYYY-MM-DD

### Attendees
Alice Chen, Tom Nguyen, Sarah Kim...

---

### Risks Raised
1. CLOUD-003 vendor contract unsigned...
   Owner: Sarah Kim. Likelihood: High.
   Impact: Critical.
   Mitigation: escalate to CPO...

2. DATA-001 had 3 pipeline failures...
   Owner: Data Eng 2. Likelihood: High.
   Mitigation: freeze upstream changes...

---

### Dependencies Discussed
- PROD-002 blocked by CLOUD-003...
- SEC-002 shares resources with CLOUD-003...

---

### Key Wins
1. PROD-001 hit 100K beta users...
2. INT-001 CI/CD pipeline 40% faster...

🔄 Auto-Update via Claude

Step 0b in run_weekly_report.py:

Before generating the report, the script calls Claude with the live Jira DataFrame to auto-generate 2 fresh items per section:

Risks Raised — 2 items from overdue/critical issues Dependencies — 2 items from blocked/linked issues Key Wins — 2 items from recently resolved/completed

The updated file is then read by both the escalation agent (context for VP email drafts) and the report generator (risk register slides).

Each entry includes: issue key, description, owner, likelihood, impact, mitigation, and dates — providing full traceability

Agent Manifest

Agent	Script	Trigger	Input	Output
Jira Monitor	`jira_monitor.py`	Daily 8:08 AM	Jira REST API snapshots	Delta email + JSON snapshot
Escalation Agent	`escalation_agent.py`	Daily (after monitor)	Jira snapshot + triggers	VP escalation email draft
Weekly Report	`run_weekly_report.py`	Weekly (manual/scheduled)	Jira CSV export	5-slide PPTX + JSON analysis

PPTX Slide Contents — 5 Slides

Cover— Title, date, project scope

Snapshot— RAG KPIs + project-level table (Red/Amber/Green counts per project)

Decisions— Escalations, blockers, items needing VP decision

Wins + Risk— Completed milestones, upcoming risks, dependency status

Narrative— Claude-generated executive summary, 3-5 sentences, VP-ready

Escalation Triggers

New issue with Highest or Critical priority detected

Issue overdue by 3+ days (past due date)

Priority escalated overnight (e.g. Medium → Highest)

3+ overdue issues in the same project area

Scheduled Tasks (Claude Code Routines)

Task	Schedule
Jira daily dummy injector	7:00 AM daily
Daily portfolio health dashboard report	8:08 AM daily
Risk and dependency report	8:00 AM Mondays

Environment & Config

API & Model

ANTHROPIC_API_KEY
Model: claude-sonnet-4-6
Anthropic Python SDK

Jira

JIRA_BASE_URL
JIRA_EMAIL
JIRA_API_TOKEN

GMAIL_USER
GMAIL_APP_PASSWORD
SMTP via Gmail

GitHub

Repo: vppuri-vjra/vp-report-generator
GitHub Actions for CI

Output Artifacts

VP_Portfolio_YYYY-MM-DD.pptx 5-slide executive deck

VP_Analysis_YYYY-MM-DD.json Full Claude analysis + RAG data

snapshots/jira_YYYY-MM-DD.json Daily Jira state snapshot

escalation_draft.json VP escalation email draft

Cost

Weekly Report

~$0.05 per run
1,500–2,500 tokens per Claude call

Daily Monitor + Escalation

~$0.02/day
Snapshot comparison + trigger check

End-to-End Flow — 8 Steps

Jira API Pull— Fetch current issues from Jira REST API (KAN project), save as JSON snapshot

Snapshot Compare— Diff today's snapshot vs yesterday's — detect new issues, status changes, priority shifts

Delta Email— Email daily change summary to stakeholders via Gmail SMTP

Escalation Check— Scan snapshot for Red/Critical triggers — if any fire, draft VP escalation email

CSV Export— Export Jira data as CSV for weekly report generation

Claude Analysis— Send CSV to Claude API for RAG classification, risk analysis, and executive narrative generation

PPTX Build— Generate 5-slide PPTX deck from Claude's structured analysis using pptxgenjs

Auto-Open / Email— Open PPTX locally or email to VP with attached deck and JSON audit trail

Slide Deck Structure — 5 Slides

SLIDE 1 · Cover

VP Portfolio Health Report

June 28, 2026

Projects: KAN, PLAT, INFRA, DATA, MOBILE

Generated by Claude Agent Pipeline

SLIDE 2 · Snapshot

G: 12 A: 5 R: 2

KAN: 8G 2A 1R
PLAT: 4G 3A 1R
INFRA: all green

SLIDE 3 · Decisions

Needs VP Decision:

• KAN-234: API migration blocked on vendor
• PLAT-89: Infra budget increase request
• DATA-156: Schema change timeline

SLIDE 4 · Wins + Risk

Wins:

• MOBILE sprint 8 completed on time
• INFRA migration 100% done

Risks:

• KAN vendor dependency
• DATA team at capacity

SLIDE 5 · Narrative

"Portfolio is largely on track with 12 of 19 epics Green. Two Red items in KAN require VP decision on vendor timeline. INFRA migration completed ahead of schedule. Recommend focusing next sprint on clearing the 5 Amber items before Q3 planning."

Escalation Email Example

To: vp-engineering@company.com

Subject: ESCALATION: 2 Critical Issues Detected — KAN Project

From: portfolio-agent@company.com (automated)

Good morning,

The daily portfolio scan detected 2 escalation triggers requiring your attention:

KAN-234 — API vendor contract expired. Priority escalated from Medium to Highest overnight. Blocking 3 downstream tasks.

KAN-267 — Security patch overdue by 4 days. Originally due June 24. Assigned to: J. Park.

Recommended action: Schedule 15-min sync with KAN lead (K. Matsumoto) to unblock vendor decision.

This email was auto-generated by the Portfolio Health Agent. Full snapshot: snapshots/jira_2026-06-28.json

Delta Report Example — Daily Change Summary

To: tpm-team@company.com

Subject: Daily Portfolio Delta — June 28, 2026

New Issues (3)

KAN-301 — Add rate limiting to public API (Medium, To Do)

PLAT-112 — Dashboard performance regression (High, In Progress)

DATA-170 — ETL pipeline timeout on large datasets (Medium, To Do)

Status Changes (5)

KAN-289 — In Progress → Done

KAN-291 — To Do → In Progress

MOBILE-45 — In Review → Done

MOBILE-48 — To Do → In Progress

INFRA-78 — In Progress → Done

Resolved (2)

KAN-289 — User authentication refactor (completed)

INFRA-78 — Database migration to v3 (completed)