Build Protocol · Decision

Apply rigorous process to decisions that cost money, time, or opportunity, before committing, not after.

Inherits the 8 Iron Rules from build-protocol. Adds 3 decision-specific rules and a 10-step workflow tuned for high-stakes choices.

Intent

Use this skill when facing a decision that commits more than $1,000 or more than 1 day of effort, is hard to reverse, or involves real tradeoffs between multiple alternatives. This includes investments (stocks, crypto, real estate), major purchases, technology selection, supplier choice, and buy-vs-build decisions. The skill enforces real-time data verification (never use stale document prices), methodology transparency (PE/DCF/decision matrix required, no gut calls), position-sizing discipline, and stop-loss planning before entry. Triggers on explicit queries like "should I buy X", "compare options", "pick a vendor", "which option", and variations in Chinese like "选哪个" or "怎么选". Do not use for reversible low-cost choices or pure knowledge production where build-protocol suffices.

Inputs

Parameters:

• Decision category (investment, purchase, technology selection, supplier choice, or other)
• Budget ceiling (hard limit in currency or time)
• Timeline (when decision must be made; when it takes effect)
• Risk tolerance (for investments: max acceptable drawdown as % or absolute amount)
• Non-negotiable constraints or hard-no's (e.g., must integrate with Salesforce, must ship by Q2)

External Connections (Real-time Data):

• Stock prices: Stooq API (https://stooq.com/q/l/?s=SYMBOL.us&f=sd2t2ohlcv&h&e=csv) or Yahoo Finance (https://query1.finance.yahoo.com/v8/finance/chart/SYMBOL); no API key required, public endpoints
• Cryptocurrency prices: CoinGecko API (https://api.coingecko.com/api/v3/simple/price?ids=bitcoin&vs_currencies=usd); no authentication required
• Product prices: direct web scrape of current retailer listing (Amazon, B2B platform, etc.); requires parsing HTML, cache-busting headers recommended
• Forex rates: Stooq (same command as stocks, e.g., s=USDJPY)
• Vendor pricing: fetch current pricing page directly from vendor domain (not cached PDFs or email quotes older than 48h)

Context / Prior Work:

• Existing comparison matrices or decision documents (if iterating on a prior analysis)
• Historical P/L logs or decision postmortems (to avoid repeating mistakes)
• Relevant market data: sector median PE ratios, historical volatility, competitor benchmarks

Edge Cases & Gotchas:

• Real-time data unavailable (e.g., market closed, API down, product delisted): state explicitly; do not substitute document prices from analysis older than 24h
• Rate limiting on public APIs: retry with exponential backoff; cache results for 1h to avoid hammering endpoints
• Stale quotes in browser cache: use curl or fresh HTTP requests with Cache-Control: no-cache headers
• Cryptocurrency slippage: live price on CoinGecko differs from actual fill on exchange; account for slippage in position sizing

Procedure

Pre-decision (Steps 1-4)

Step 1: Research and gather options

Inputs: decision category, budget, constraints

Process:

• List all viable candidates (minimum 3 unless disqualified); do not anchor on first option or incumbent
• Source from web search, product review sites (TrustRadius, G2 for SaaS; Amazon for consumer goods), market data (Yahoo Finance, Crunchbase for companies), expert opinions (Gartner reports, analyst notes)
• For each option, collect: name, price/cost, key features, availability, lead time

Outputs: options list (text or table), 1-2 sentences per option describing scope and primary use case

---

Step 2: Set constraints and hard limits

Inputs: business/personal requirements, available budget, timeline

Process:

• Define budget ceiling (hard limit; exceeding it is disqualifying)
• Define timeline (when decision deadline is; when implementation begins)
• List hard-no's (non-negotiable disqualifiers: e.g., "must be GDPR-compliant", "must support our legacy system", "supplier must have US tax ID")
• For investments: state max acceptable drawdown (% or $) you can stomach without panic-selling

Outputs: constraints document (can be bullet list); explicitly state what is negotiable vs non-negotiable

---

Step 3: Fetch real-time data (non-skippable)

Inputs: list of options with identifiers (stock ticker, product SKU, vendor name, etc.)

Process:

• For stocks: run curl command to Stooq or Yahoo Finance; extract open, close, 52-week high/low, volume
• For crypto: call CoinGecko API; record spot price in your preferred fiat currency, 24h high/low, market cap
• For products: navigate to retailer website (not cached email quote); record current list price, promotional discount (if any), estimated delivery
• For services/vendors: visit vendor pricing page; record per-unit cost, volume discounts, contract terms
• Mandatory check: if data is older than 24h, refetch or state "data refreshed [timestamp]"
• If live data unavailable: state explicitly ("stock halted", "product out of stock", "vendor pricing page unreachable"); do not fall back to document prices

Outputs: real-time price snapshot with timestamp (e.g., "BTC: $43,250 @ 2026-04-30 14:22 UTC"), formatted as table or structured text

---

Step 4: Build comparison matrix

Inputs: options list, real-time data, constraints

Process:

• Define 4-8 scoring dimensions relevant to this decision type:
  • For investments: valuation (PE, Price/Book), momentum (52w return, MA50/MA200 cross), volatility, dividend yield, sector exposure
  • For purchases: upfront cost, total cost of ownership (TCO) over 3 years, reliability/warranty, resale value, time-to-value
  • For technology: initial setup cost, monthly/annual fees, integrations available, vendor stability, user reviews (weighted by recency)
  • For suppliers: unit price, minimum order, lead time, quality rating, contract flexibility
• Score each option on each dimension (1-5 scale, percentile ranking, or weighted score)
• Make tradeoffs explicit: identify which dimensions conflict (e.g., "lowest cost" vs "fastest delivery")

Outputs: comparison matrix (table or spreadsheet) with options as rows, dimensions as columns, and scores filled in; add a "notes" column explaining any score that is not self-evident

---

Decision (Steps 5-7)

Step 5: Apply methodology (show your work)

Inputs: comparison matrix, real-time data, constraints, decision category

Process:

• For investments:
  • Calculate PE ratio: (current price) ÷ (annual earnings per share); compare to sector median PE
  • Run DCF (discounted cash flow): project 5-year cash flows, apply 8-12% discount rate (adjust for risk), sum terminal value; show all inputs and discount rate sensitivity (run 2-3 scenarios with different discount rates)
  • Check technical levels: 50-day moving average (MA50), 200-day moving average (MA200); is price above/below? (lagging indicator; context only, not prediction)
  • Identify catalyst: what event would validate/invalidate this thesis? (earnings announcement, product launch, regulatory decision, etc.)
• For purchases:
  • Calculate TCO: upfront cost + (annual operating cost × 3 years) - (residual value at year 3); include maintenance, energy, consumables
  • Compare TCO per unit of benefit (e.g., TCO ÷ throughput, TCO ÷ seats)
• For technology/vendor:
  • Use weighted decision matrix: assign weights to each dimension (e.g., "cost = 30%, security = 40%, ease-of-use = 20%"), multiply each option's score by weight, sum per option
  • Benchmark against analyst reports (Gartner, Forrester quadrants) if available
• For all decisions: show every calculation; if you used an estimate (e.g., "assume 3% annual growth"), state it explicitly

Outputs: methodology document with all inputs, calculations, and assumptions documented; show sensitivity analysis if using DCF or complex models; append "(est.)" to any number derived from an estimate or source older than 24h

---

Step 6: Plan exit conditions and Plan B

Inputs: chosen option (tentative), constraints, risk profile

Process:

• For investments:
  • Define hard stop-loss: at what price or % loss do you exit? (e.g., "exit if price drops >15% from entry")
  • Define thesis-break trigger: what would prove your analysis wrong? (e.g., "if company misses earnings by >20%", "if short interest spikes above 30%")
  • Set profit target (secondary): at what % gain do you take profits? (less important than stop-loss, but plan it)
• For purchases:
  • Confirm return policy window (days to return with refund); resale value estimate if you bail out early
  • Backup plan: if this option fails, which alternative is second choice? (and can you switch to it within timeline?)
• For technology/vendor:
  • Review contract termination terms: notice period, early-exit penalties, data export/migration requirements
  • Estimate migration cost and effort if you need to switch vendors mid-contract
• Write these down before committing. Do not plan exits after things go wrong.

Outputs: exit plan document (plain text or structured format) with stop-loss level, thesis-break trigger, profit target (for investments), and fallback option clearly stated

---

Step 7: Make decision and document rationale

Inputs: comparison matrix, methodology, exit plan, constraints

Process:

• State chosen option clearly and unambiguously (e.g., "Decision: Buy Acme CRM Pro")
• Write "Why This Option" (one paragraph): explain how this choice fits your constraints, risk tolerance, and timeline; do not list why others are bad
• Write "Why Not the Alternatives" (one paragraph): for the top 2-3 runners-up, explain the specific reason you chose the winner instead; be honest, not dismissive
• Include a date-stamp and note the assumptions that underpin this decision (they will change; you need to know how old this reasoning is)
• Explicitly list any assumptions that could break the thesis (e.g., "assumes NVIDIA continues to lead AI chip market", "assumes AWS pricing remains stable")

Outputs: decision statement (2-3 paragraphs, dated, with methodology reference and key assumptions listed)

---

Post-decision (Steps 8-10)

Step 8: Execute with position sizing

Inputs: decision statement, exit plan, real-time entry price, account size (for investments)

Process:

• For investments:
  • Calculate position size: (account risk per trade in %) ÷ (entry price - stop-loss price)
  • Example: if account = $10k, risk = 1% ($100), entry = $50, stop-loss = $45, then size = $100 ÷ $5 = 20 shares
  • Do not max-leverage on first entry; position size conservatively (1-2% of account per position)
• For purchases:
  • Confirm budget compliance before checkout (do not exceed ceiling)
  • If multi-stage (e.g., pilot before rollout), phase entry: do not commit 100% on day one
• For technology/vendor:
  • Phased rollout: pilot with 1 team/region before company-wide deployment
  • Document entry conditions: who owns the implementation? what are success criteria for pilot phase?

Outputs: execution order (purchase order, trade confirmation, contract signature) with position size, entry price (real-time), stop-loss, and profit target clearly documented

---

Step 9: Track performance daily (not weekly, not monthly)

Inputs: entry price, position size, daily market data

Process:

• Record daily log with: date, current price, unrealized P/L (in $ and %), reason for any major price moves (news, earnings, sector rotation, etc.)
• For non-investment decisions: track KPIs that validate the decision (tool adoption rate, cost savings vs projection, integration success rate, time-to-value metrics)
• Daily tracking catches thesis-breaking moves early; weekly/monthly is too slow
• If thesis is broken before stop-loss is hit, consider exiting early (e.g., company announces bankruptcy, product features are not working, vendor loses major contract)

Outputs: daily log (spreadsheet or text file), updated each day, with at least: date, price, P/L %, notes on events that affect thesis

---

Step 10: Review and adjust; kill bad bets fast

Inputs: daily log, original methodology, exit plan, current market conditions

Process:

• Weekly review: is thesis still intact? Did any catalyst happen? Did stop-loss get hit? Are there new facts that contradict your entry analysis?
• Monthly review: compare actual outcome vs pre-decision projection; document unexpected changes
• If thesis is broken: exit at stop-loss immediately; do not average down hoping it recovers (throwing good money after bad)
• If thesis is intact but challenged: review the specific challenge against your original methodology; decide whether to hold or reduce position
• Log lessons in a decision postmortem: even for winners, document why it worked (so you repeat it); for losers, document what you missed
• Use postmortems to improve your process (e.g., "I should have checked vendor reviews more carefully", "my DCF discount rate was too optimistic")

Outputs: weekly review notes, monthly postmortem (one per decision, even winners), updated decision log in central tracking system

---

Decision Points

Decision Point 1: Data freshness on real-time data fetch (Step 3)

If live data is available (market open, API responding, product in stock):

• Use the live price as entry point for all downstream calculations
• Timestamp it and note data source

Else (market closed, API down, product delisted, vendor pricing page unreachable):

• Do not substitute a price from a cached document, email, or analysis older than 24h
• State explicitly in the methodology document: "Real-time data unavailable; decision deferred until [data source is available]"
• If decision is urgent and you must proceed without live data, disclose this risk and flag for post-decision audit

Decision Point 2: Methodology presence check (Step 5)

If you have applied a quantitative methodology (PE ratio, DCF, weighted matrix, TCO calculation):

• Proceed to Step 6 with confidence; your analysis is defensible

Else (you have only gut feelings, anecdotal evidence, or "it sounds good"):

• Do not proceed to Step 7 (decision)
• Return to Step 5 and apply a methodology appropriate to the decision category; use templates in references if needed
• "I think this is a good buy" is not a methodology; it is a guess

Decision Point 3: Exit plan completeness (Step 6)

For investments:

• If stop-loss and thesis-break trigger are defined: proceed to Step 7

Else:

• Do not commit capital; return to Step 6 and define these before entry
• "I will decide later" = rationalizing losses when they happen; exits must be pre-planned

For purchases and technology selections:

• If fallback option and return/termination policy are documented: proceed to Step 7

Else:

• Document these before committing; incomplete exit planning = being trapped in a bad choice

Decision Point 4: Sycophancy of precision (Step 5 and 7)

When outputting numbers derived from estimates, multi-source blends, or stale data:

• If the number is an estimate: append "(est.)" or "(含估算)"
• If the number is from a source older than 24h: append the source date, e.g., "(data from 2026-04-28)"
• If the number is compounded from multiple uncertain inputs (e.g., DCF with estimated growth rate): round to 2 significant figures, not 2 decimal places
• Example: DCF estimate of "12.85%" is sycophancy; "~13% (est., based on 2-8% growth range)" is honest

Do not output false precision. Decimal places do not add credibility when the underlying inputs are estimates.

Decision Point 5: Post-decision thesis break (Step 10)

During daily/weekly review, if evidence emerges that contradicts a key assumption from Step 5:

• Example: you chose based on "company has strong cash position", but then company announces layoffs / dividend cut / acquisition at loss
• Example: you chose a tool based on "integrates with Salesforce", but integration is buggy and team is blocked
• If thesis is broken: exit or reduce position immediately; do not average down
• If thesis is intact but questioned: review the specific challenge; quantify the impact (1% of thesis, 20%, 50%?); decide whether to hold or exit

Output Contract

Step 1 output (options list):

• Format: bullet list or 1-column table
• Fields: option name, price/cost, 1-2 sentence description
• Minimum 3 options listed

Step 2 output (constraints document):

• Format: bullet list or structured text (YAML, markdown, plain text)
• Fields: budget ceiling, timeline (decision deadline + implementation start), hard-no's, risk tolerance (for investments)
• All constraints must be quantified or explicitly defined; "moderate budget" is not acceptable

Step 3 output (real-time data snapshot):

• Format: table or structured list (JSON, CSV, plain text)
• Fields: option name, current price, timestamp (ISO 8601), data source (URL or API name), 24h high/low (for liquid assets)
• Must include timestamp; data older than 24h is flagged as stale

Step 4 output (comparison matrix):

• Format: table (spreadsheet, markdown table