Engineering Strategic Prompt Chains
Given an existing artifact (an audit, a framework, an inventory, a strategy doc), produce a follow-up prompt that:
- Names the deliverable explicitly (Title Case, verb-first: "Build the X Matrix", "Design a Y Framework")
- Opens by referencing the specific prior artifact by name and what it already covers
- Identifies the gap — what the prior artifact describes but doesn't operationalize
- Specifies a numbered list of concrete components the new deliverable must contain (usually 3-5)
- Anchors every component to real specifics from the source material (numbers, names, thresholds — never generic placeholders)
- Closes with a one-sentence payoff explaining why this deliverable matters strategically
Example skeleton:
[Verb] a [deliverable name] that [operationalizes/extends/stress-tests] the
[named prior artifact]. This [deliverable] should [provide/map/develop]:
(1) [specific component tied to source detail],
(2) [specific component tied to source detail],
(3) [specific component tied to source detail],
(4) [specific component tied to source detail].
[Closing sentence: why this transforms/strengthens/validates the broader effort.]
Progress:
- Step 1: Identify the source artifact and extract its concrete specifics (numbers, entity names, thresholds, categories)
- Step 2: Determine the gap class — is the prior artifact descriptive/diagnostic but missing an execution plan, a scoring system, a prioritization matrix, or a validation layer?
- Step 3: Choose a deliverable type that fills that gap (see Deliverable Types below)
- Step 4: Draft the opening sentence naming the source artifact and its known scope
- Step 5: Build the numbered component list, ensuring each item references a real specific from the source (not a generic category)
- Step 6: Write the closing payoff sentence tying the new deliverable back to the larger goal (investor confidence, data integrity, competitive advantage, etc.)
- Step 7: Title the prompt as a short, verb-first, Title Case action phrase
Deliverable Types (pick based on the gap)
- Execution plan — when a framework exists but hasn't been applied step-by-step to real data
- Scoring/decision matrix — when criteria exist but lack weighting, thresholds, or comparative ranking
- Impact/risk assessment — when an action (merge, migration, launch) is planned but consequences aren't mapped
- Competitive scorecard — when gaps vs. competitors are identified but not quantified side-by-side
- Roadmap/pathway — when constraints are fixed but growth/upgrade scenarios aren't modeled
- Journey/touchpoint map — when a diagnostic exists but hasn't been converted into intervention points
Example 1: Input: A dataset audit found 74 artifacts with possible duplicates based on title/content similarity, no merge rules yet exist. Output: "Develop Deduplication Rules and Merge Strategy" — a framework defining matching criteria (true duplicate vs. intentional variant), a retention prioritization system, and a merge/cross-reference-preservation strategy, closing with why this protects the Corpus-to-Volume Traceability Matrix.
Example 2: Input: A competitive audit lists specific friction points parents hit on Zillow/Redfin/Realtor.com across 4 journey stages, with a stated 85,000 monthly high-intent queries figure. Output: "Build Competitive Friction Point Scorecard Matrix" — a 1-5 scored matrix across the 4 friction points x 3 competitors, with a column for the proposed platform's advantage, closing with why this validates capturing the 85,000 queries.
Example 3: Input: A business model has fixed constraints ($25/month budget, 1,000-row Airtable cap, 500-525m spatial accuracy). Output: "Design a Constraint Negotiation & Upgrade Pathway" — a tiered roadmap showing what triggers a move to $50/$100 budgets, a database migration, or a premium accuracy tier, closing with reframing constraints as pricing/segmentation strategy.
- Always chain, never restart. Every new prompt must explicitly cite the prior deliverable by name so the sequence reads as a continuous escalation, not a disconnected request.
- Use real numbers from context. "74 artifacts," "$25/month," "500-525 points," "85,000 monthly queries" — specificity signals the chain is building on itself, not templating.
- Escalate concreteness. Each link in the chain should move from strategy → framework → execution plan → scoring/validation → roadmap. Don't loop back to strategy once you've reached execution.
- Keep components mutually exclusive. The 3-5 numbered items in each prompt should cover distinct facets (e.g., criteria, prioritization, merge mechanics, validation, logging) not overlapping restatements.
- End with stakes, not summary. The closing sentence should state a consequence (risk avoided, advantage gained, confidence built) rather than just recapping the list.
- Title as an imperative action. "Build," "Design," "Develop," "Apply," "Map," "Execute" — never a noun phrase alone.
- Do not introduce a deliverable type disconnected from the source artifact's actual gap (e.g., asking for a "scoring matrix" when the real gap is a missing execution sequence).
- Do not use placeholder specifics ("Artifact A vs Artifact B") when the source material has real names, counts, or thresholds available — always substitute the real detail.
- Do not stack more than 5 numbered components; beyond that, split into two chained prompts instead.
- Do not omit the reference to the originating artifact — a prompt that could stand alone without context is not part of the chain and defeats the methodology's purpose.
- Do not end on a flat recap ("This will help organize the data") — always state the strategic stake.