WorkBuddy Practical: 18 - Portfolio Management & Investment Analysis

Portfolio Management Skill Stack

• stock-advisor (Core): Analyzes charts, reads filings, cross-validates data, runs advisory group debates, and formats reports.
• a-share-analyst: Tracks market indices, computes technical levels, and runs daily screeners.
• financial-expert: Queries financial databases, filters stocks, and fetches research PDFs.
• peers-advisory-group: Runs multi-agent debates with specialized roles to test ideas.

Use a-share-analyst and financial-expert for daily screening, stock-advisor for deep research on individual stocks, and peers-advisory-group to test assumptions.

Reusable Prompts for Investment Pipelines

Replace placeholder tags 【】 with your target stock or code.

Prompt 1: Compiling a Factual Foundation

Analyze the baseline profile of [Target Stock/Code] and compile a structured summary:
1) Core business tracks and product portfolios.
2) Revenue streams and margin allocations.
3) Primary customers and application scenarios.
4) Position and value capture in the supply chain.
5) Key strategic transitions over the past 3 years.

Rules:
- Use only verifiable primary disclosures.
- Map 3-5 bullet points per section.
- Compile facts only; do not suggest buy/sell actions.

Prompt 2: Analyzing Industry Dynamics

Analyze the industry landscape for [Target Stock] in [Target Sector] from a research perspective:
1) Industry cycle stage (recovery, expansion, decline, stagnation) backed by capex and inventories.
2) Supply-demand metrics: industry capacity utilization, inventories, and delivery lead times.
3) Price dynamics: index histories, spreads, and cost pass-through capacities.
4) Competitor concentrations (CR5) and market shares.
5) Key external factors (interest rates, trade tariffs, state subsidies).
Differentiate long-term structural changes from short-term volatility. Output: cycle stage judgment, key metrics tables, and 3 leading / 3 lagging indicators.

Prompt 3: Business Model Analysis (Business Engine)

Analyze the business engine of [Target Stock] from a fundamental perspective, focusing on:
How does this company make sustainable profits over the long term?

Rules:
- Rely strictly on verifiable filings (annual reports, prospectus, regulatory filings).
- Highlight facts separately from opinions.
- Output a structured Markdown report.

Required Sections:
1. Core business logic in a single sentence (under 50 words: what it sells -> to whom -> why it is profitable).
2. Quantified business structure: revenue split, margins, and 3-5 year growth trends for each segment. Define profit engines vs. high-revenue, low-margin segments.
3. Revenue mechanics: transactional vs. subscription models, cost structures (raw materials, R&D, sales), margin drivers, and scale efficiencies.
4. Customer concentration (Top 5 / Top 10 percentages), sales channels (direct vs. distributor), pricing power (historical price increases), and customer switching costs.
5. Subsidiary contributions: log key subsidiaries, operational sustainability, and split of recurring vs. non-recurring profits (subsidies, asset sales).
6. Business vulnerabilities: identify dependencies, failure points, competitor threats, and list 3-5 metrics to monitor.

Output:
- Business engine summary.
- Segment breakdown table (revenue, margin, growth).
- Revenue engine logic flow.
- 3 key conclusions for long-term investors.

Prompt 4: Auditing Financial Quality

Analyze the financial quality of [Target Stock] over the past 3-5 years:
1) Alignment of revenue, net profit, and operating cash flows.
2) Changes in accounts receivable, inventories, and contract assets.
3) Contribution of non-operating items to profits.
4) Impact of one-off write-downs or changes in accounting policies.
5) Crucial financial warning signs.

Rules:
- Perform "Net Profit vs. Operating Cash Flow" validation.
- Provide explanations for anomalies.
- Highlight metrics to track.

Prompt 5: Governance and Insider Transactions

1. Analyze the governance structure of [Target Stock]:
   - Voting control, major shareholders, board compositions.
   - Share pledges, insider selling schedules, control transition risks.
   - Related party transactions and horizontal competition.
   Output: governance map, risk matrix, and watchlist logs.
2. Compile a capital timeline for the next 12 months: lock-up expirations, share buyback schedules, private placement updates. Assess potential market supply impacts.
3. Review management incentives: verify performance targets (revenue vs. margins vs. ROIC), target difficulties, and potential short-term manipulation risks.

Prompt 6: Mapping Market Discrepancies

Compile market discrepancies regarding [Target Stock]:
1) Core arguments of the bull case.
2) Core arguments of the bear case.
3) Key metrics supporting each side.
4) How these differences can be validated by upcoming data releases.
5) Timelines for verification.
Remain neutral; do not take sides or use subjective phrasing.

Prompt 7: Moat Strength and DCF Auditing

Analyze the moat strength of [Target Stock]:
1) Pricing power: gross margins and cost pass-through history over 5-10 years.
2) Switching costs: systems integration, compliance barriers, or ecosystem locks.
3) Cost/scale advantages: cost advantages driven by scale or distribution.
4) Intangible assets: brand premiums, patents, licenses, and proprietary databases.
5) Competitor actions: history of competitor responses and defense actions.
Output: Moat rating (0-5), validation table, erosion risks, and metrics to monitor.

Build a DCF model for [Target Stock] using public metrics:
- Define and justify WACC and terminal growth rates.
- Project free cash flows for 5-10 years: revenue growth, margins, capex, and reinvestment rates.
- Output a sensitivity matrix (WACC vs. terminal growth, WACC vs. operating margin).
- Reverse DCF: calculate what revenue and margin path the current stock price implies.
Output: valuation range, core assumptions table, and key sensitivity flags.

Prompt 8: Complete DeepResearch Template

I need a comprehensive Investment Due Diligence Report for [Target Stock/Code] under [Investment Style] for a holding horizon of [Timeframe].

Constraints & Standards:
1. Cover 3-5 years of historical financial data (CAGR) and WACC/multiples over 5-10 years.
2. Fact-first: Cite annual reports, prospectuses, and regulatory inquiries. Differentiate facts from opinions.
3. Cross-validate: run profit vs. cash flow audits and competitor comparisons.
4. Adversarial: Include bear logic and tail risk analysis.

Workflow:
Phase 1: Business Engine & Moat
1. Segment metrics: revenue split, margins, and subsidiary contributions. Differentiate core operating profits from investment returns or state subsidies.
2. Moat validation: gross margin trends, pricing power, customer retention, switching costs, and market share ceiling.
Phase 2: Industry Context
1. Cycle mapping: define cycle stage (recovery, peak, contraction, trough) using inventory cycles and capex trends.
2. Supply-demand: track industry capacity growth, lead times, and pricing indexes.
3. Competition: market concentration shifts and competitor strategies.
Phase 3: Financial Quality Audit
1. Baseline: calculate revenue CAGR, profit CAGR, ROE (DuPont analysis), and margin trends.
2. Warning signs: inventory builds, accounts receivable days, deductions of non-recurring items, and net profit vs. operating cash flows.
Phase 4: Governance & Capital Allocation
1. Shareholder analysis: controlling interests, pledges, insider selling logs.
2. Allocation: review major acquisitions, share buyback values, dividend payout ratios, and ROIC.
Phase 5: Valuations & Risks
1. Multiples: PE/PB/PS history and peer comparisons.
2. Reverse DCF: reverse-engineer the growth rate and margin trajectory implied by the current market cap.
3. Bear cases: analyze short reports, bearish notes, regulatory threats, and tail risks.

Output Structure:
1. Investment Summary (Rating: Buy/Hold/Sell, and Core Thesis).
2. Financial metrics table.
3. Detailed analysis by Phase.
4. Valuation dashboard.
5. Watchlist monitoring checklist (positive triggers vs. exit rules).
6. Primary source bibliography.

Instead of running these prompts manually and transferring intermediate outputs, use the stock-advisor Skill to automate the pipeline.

System Design of stock-advisor

The stock-advisor Skill orchestrates existing tools into a unified pipeline:

User Input (chart or code) -> Technical levels -> Financial audit -> Cross-validation -> Multi-agent debate -> Formatted report

The Skill asks for user preferences (risk index, timeframe, sectors) during first-time configuration, and runs both image-based and code-based inputs through the same pipeline.

Case Study: Analyzing TF Communication (300394)

The stock-advisor processes the research in three steps:

Step 1: Chart Analysis (Technical Read)

Prompt:

Analyze the daily chart and MACD metrics for A-share 300394:
1. Identify stock name and current price.
2. Analyze candlestick pattern over the last 5 days.
3. Check moving average arrangements (MA5, MA10, MA20) for crossover alerts.
4. Check MACD line positions and signal changes.
Output technical summary table.

Output:

• Trend: MA5 > MA10 > MA20, standard bullish alignment.
• Risk Flag: Long upper shadow on current bar, MACD histogram starting to contract.
• Levels: Support at MA5 (347) / MA10 (319), resistance at 376.

Step 2: Financial Audit

Prompt:

Incorporate Q3 financial data and profit forecasts for A-share 300394.
Provide:
- Technical summary.
- Financial audit (revenue growth, margins, valuations).
- Volume trend.
- Rating (Strong Buy/Buy/Neutral/Underperform/Sell).
- Operation suggestions for short-term (1-2 weeks) and mid-term (1-3 months).
- Support and resistance levels.

Output: Pulls metrics (revenue 3.918B, +63.63%, net profit 1.465B, ROE 31.30%, margin 51.87%, PE 146.70) and calculates the加权 weighted rating:

Step 3: Multi-agent Debate (Advisory Group)

Prompt:

I am undecided. Run a multi-agent debate regarding 300394:
- Warren Buffett: Value investing (intrinsic values, margins of safety).
- Elon Musk: Tech disruption and innovation.
- Bill Gates: Business models and competitive landscapes.
- Steve Jobs: Product design and user experience.
Debate rules:
1. Each agent presents a 3-minute argument.
2. Agents cross-examine each other's points.
3. Each agent provides a Buy/Hold/Sell recommendation.
4. The host synthesizes a final action plan.
Incorporate current financial metrics.

Output: Pulls real-time metrics (full year revenue 5.163B, profit 2.017B, Q1 quarter-on-quarter contraction, competitor metrics for Zhongji Innolight) to feed the debate.

• Buffett: PE of 142 offers zero margin of safety. Hold off until valuation corrects below 200.
• Gates: Strong fundamentals but competitor valuations are more attractive. Wait for multiple contraction.
• Musk: CPO is the optical transition layer. High valuations represent scaling growth. Buy.
• Jobs: Keep holding if you believe in the CPO architecture, but verify that specifications are met by 2026. Synthesized plan: Hold current positions, do not add capital at peak valuations, and set a buying trigger at 60x PE.

Tencent ima Knowledge Base Integration

See the detailed guidelines in Chapter 16 to sync stock-advisor portfolios with ima. This secures RAG checks against corporate disclosures, and keeps track logs of financial reports.

Detailed Walkthroughs of the 13 Case Submissions

The submissions/ directory logs 13 operational case files:

• daily-ai-news: news aggregator schedules.
• portfolio-tracker: portfolios logs.
• report-builder: documents compilers. Ensure files are cataloged in correct workspace directories, and verify that TraceIDs are logged in Cube Sandbox to maintain compliance.

Prompt 1 requirements

• Use only verifiable information;
• Give three to five bullet points for each section;
• Organize facts only and do not provide investment advice.

Prompt 2 requirements

Separate long-term structural factors from short-term volatility, and output a cycle-stage judgment, an evidence-chart checklist, and three leading and three lagging indicators.

Prompt 3 requirements and required structure

Use annual reports, prospectuses, periodic announcements, investor-exchange minutes, and authoritative industry reports. Separate Fact from Judgment, and support every judgment with evidence or a logic chain.

The report must include:

1. A one-sentence explanation, in no more than 50 words, of what the company sells, to whom, and why it can make money;
2. A quantified breakdown of every business or product line, including revenue share, gross margin, and three-to-five-year growth, plus a judgment about which lines generate profit and which have high revenue but little or no profit;
3. The revenue engine for each core business: how money is collected (one-off, subscription, repeat purchase, or project), main costs, margin drivers, and scale effects;
4. Customers, channels, Top 5/Top 10 concentration, pricing power, price increases, and switching costs;
5. Key subsidiaries, associates, and non-recurring operations, separating sustainable operations, cyclical gains, investment income, subsidies, and asset sales;
6. Business stability and fragility, including broken assumptions, competitive, technology, or policy shocks, and three to five monitoring metrics.

Prompt 3 final output

Output a one-sentence business-essence conclusion, a revenue/profit/margin table, a written revenue-engine logic chain, and the three most important conclusions for a long-term investor.

Prompt 4 research principles

Do not predict the stock price. Judge financial quality only. Force a “net profit versus operating cash flow” cross-check, explain every anomaly with a hypothesis and a verification path, and identify the metrics worth tracking continuously.

Prompt 5: Governance and alignment

The purpose is to identify the risk of a good business with poor governance. Review the controlling shareholder, actual controller, board structure, pledge ratio and changes, selling plans, control-transition risk, related-party transactions, horizontal competition, and fund occupation. Output a governance map, high/medium/low risk radar, and an announcement watchlist.

Build a 12-month “share-supply event timeline” covering lock-up expirations, employee-share unlocks, private placements, rights issues, and buyback progress. For every event, assess potential selling pressure or absorption, the path to the valuation center, and historical price reactions when available.

Review management compensation and equity incentives: whether revenue, profit, cash flow, or ROIC targets can be achieved through accounting or short-term actions, how difficult the targets are compared with peers, and whether they encourage sacrificing R&D or cash generation. Output an alignment conclusion, key-term excerpts, and improvement suggestions.

Prompt 6 analysis requirements

Remain neutral. Do not take sides, give investment advice, or use emotional language. Every judgment must be testable by a future data release or event. The output must identify the bull logic, bear logic, strongest evidence for each, disagreements that can be validated, and the exact verification checkpoints.

Prompt 7: Moat evidence and DCF assumptions

For moat analysis, cite company disclosures or authoritative sources and examine pricing power over the past five to ten years, customer switching costs in systems, processes, compliance, and ecosystems, network or scale effects, brands, patents, licenses, data and channels, and how competitors have attacked or been resisted. Output a 0–5 moat score, evidence table, erosion risks, and monitoring metrics.

For the DCF, state the WACC or discount-rate assumptions and their basis, project five to ten years of free cash flow with revenue, margins, and reinvestment, provide a discount-rate-by-terminal-growth or discount-rate-by-margin sensitivity table, and reverse the current market capitalization into an implied revenue-growth and margin path. Output the valuation range, assumptions, and the two assumptions most likely to be wrong with verification plans.

Constraints and standards for the due-diligence report

1. Cover three to five years of financial trends using CAGR, and five to ten years of historical valuation ranges;
2. Separate Fact and Opinion and support opinions with annual reports, prospectuses, and regulatory inquiries;
3. Cross-validate profit against cash flow and the company against peers;
4. Include bear logic and black-swan scenarios to counter confirmation bias.

Research context

• Target: enter the stock name and code;
• Investment style: for example, value, growth, or turnaround;
• Holding period: for example, one to three years.

Workflow

Phase 1: Business engine and moat

Decompose revenue and profit, check whether the main business attracts attention while investments or subsidies generate profit, and trace subsidiaries and associates to identify drags and hidden engines. Assess pricing power, switching costs, network or scale effects, intangible assets, TAM, market share, and growth ceiling.

Phase 2: Industry cycle and supply-demand

Classify the cycle as recovery, overheating, stagflation, recession, or depression using inventory, utilization, and Capex evidence. Identify leading and lagging indicators, future capacity additions, CR5 concentration changes, price wars, and technology breakthroughs.

Phase 3: Financial health and warning signs

Calculate three-to-five-year revenue CAGR and net-profit CAGR, explain ROE with DuPont analysis, and chart gross and net margin trends. Check inventory turnover, accounts-receivable days, operating-cash-flow-to-profit, and non-recurring gains after deductions.

Phase 4: Governance and capital allocation

Review two years of issuances, buybacks, equity incentives, and major acquisitions; judge whether they accrete EPS or dilute minority holders. Check the controller's stake, high pledge risk, continuing sales by funds or executives, consistency between statements and actions, and whether historical earnings went to expansion, dividends, buybacks, or poor investments. Assess ROIC.

Phase 5: Valuation and risk

Compare current PE, PB, and PS with five-to-ten-year historical percentiles and peer valuations. Use reverse DCF to infer the three-to-five-year profit growth implied by the current price. Search for bear reports and negative sentiment, and analyze regulatory, technology-path, and geopolitical tail risks.

Output format

End the structured report with a source list and include:

1. An investment summary with a green/yellow/red rating and a one-line thesis;
2. A key financial-data table with CAGR, ROE, and cash-flow matching;
3. Detailed analysis organized by the five phases, with data attached to each conclusion;
4. A valuation dashboard showing historical percentiles, implied expectations, and peers;
5. A monitoring list: only event A strengthens the buy logic; if metric B deteriorates, such as margin below X%, the thesis is falsified and the position exits.

What problem does stock-advisor solve?

The individual prompts are useful, but a full stock study requires manually running eight prompts, copying intermediate conclusions, and formatting the report. stock-advisor turns this half-day manual chain into one orchestrated run: provide a screenshot or stock code and it retrieves data, performs technical and fundamental analysis, cross-validates sources, runs an adversarial debate, and formats the report. Human work shifts from moving text around to making and challenging the decision.

Design principle: orchestration, not reinvention

stock-advisor reuses proven parts rather than rebuilding everything. The technical scripts come from a-share-analyst, the debate from peers-advisory-group, the layout from magazine-layout, and the upload path from lark-doc. It adds only the missing fundamental-analysis and HTML-to-PDF pieces.

The productized design also records risk preference, investment horizon, sectors, and position limits on first use. Screenshot input uses image recognition plus data validation; a stock code uses a data-driven path. Both converge on the same later stages.

What the case demonstrates

The full study of TF Communication (300394) proceeds as chart → filings → advisory debate → report. The technical pass read MA5 > MA10 > MA20, a long upper shadow, a contracting MACD histogram, and support at 347/319 with resistance at 376. The financial pass extracted revenue of CNY 3.918 billion (+63.63%), net profit of CNY 1.465 billion, ROE of 31.30%, gross margin of 51.87%, and PE of 146.70, then produced a 3.65/5.0 weighted score.

The advisory group then brought in full-year revenue of CNY 5.163 billion, profit of CNY 2.017 billion, a quarter-on-quarter decline in 2026 Q1, and peer comparisons. Buffett emphasized the zero margin of safety at PE 142; Gates preferred waiting or switching; Musk argued that CPO growth justified the valuation; Jobs required the CPO FAU plan to be delivered in the second half of 2026. The host mapped the decision to investor type and checkpoints rather than issuing one universal “buy” or “sell.”

Common errors and boundaries

Do not ask AI for exact buy or sell points, trust unverified screenshot numbers, expect it to call an industry bottom, consume only the bull case, or treat an AI report as investment advice. Verify important figures against current filings, force a bear case, and keep the final decision and risk with the investor. The chapter's prompts, Skills, and case are for research assistance only and do not constitute investment advice.

Deep RAG integration with the ima knowledge base

WorkBuddy can connect to Tencent's ima knowledge base for local knowledge Q&A and research analysis:

1. Import PDFs, DOCX, XLSX, PPTX, TXT, MD files, and web links into an ima personal library;
2. Address a private library explicitly in chat with @library-name, such as @Industry Research;
3. Let the Agent retrieve matching ima passages first and then generate the brief, so financial data and metrics retain clear source support.