MOD-026-S04

Explanation Component

This submodule implements the explanation step needed to score and order candidate items against explicit goals, preferences, constraints, and diversity requirements. It returns an inspectable intermediate result for Recommendation and Ranking.

Workbook component extraction

Focused task

This submodule implements the explanation step needed to score and order candidate items against explicit goals, preferences, constraints, and diversity requirements. It returns an inspectable intermediate result for Recommendation and Ranking.

This component belongs to MOD-026 — Recommendation and Ranking. It is not an independent pattern.

Prompt overview

Complete only the explanation component task within Recommendation and Ranking. Work from the authorised candidate set, selection criteria, and user history. This submodule implements the explanation step needed to score and order candidate items against explicit goals, preferences, constraints, and diversity requirements. It returns an inspectable intermediate result for Recommendation and Ranking. Return a clearly labelled result for validation and hand-off.

Prompt templates

Replace every placeholder with authorised project information. Inspect sources, assumptions, output and human decisions before acting.

Frontier model prompt

GOAL
Help <INSERT PERSON OR TEAM HERE> achieve <INSERT PROJECT GOAL HERE> by completing the "explanation component" component of MOD-026 Recommendation and Ranking.

AIM
Produce <INSERT REQUIRED COMPONENT RESULT HERE> by completing one bounded task: the explanation component implements the explanation step needed to score and order candidate items against explicit goals, preferences, constraints, and diversity requirements. It returns an inspectable intermediate result for Recommendation and Ranking. This supports Recommendation and Ranking but does not perform or approve the complete workflow.

ROLE
Act as an experienced information-retrieval specialist. Work from the supplied material and authorised sources. Do not invent facts, records, quotations, references, permissions or results. Clearly separate source evidence, interpretation, assumptions and recommendations.

CONTEXT AND INPUTS
- Project or artefact: <INSERT PROJECT, DOCUMENT OR ARTEFACT NAME HERE>
- Intended audience or user: <INSERT AUDIENCE HERE>
- Project goal: <INSERT PROJECT GOAL HERE>
- Material to analyse or process: <INSERT OR ATTACH SOURCE MATERIAL HERE>
- Required module inputs (option set, selection criteria, and user history): <INSERT EACH REQUIRED INPUT HERE>
- Authoritative evidence or records: <INSERT APPROVED SOURCES HERE>
- Constraints, exclusions and permissions: <INSERT CONSTRAINTS HERE>
- Success criteria: <INSERT MEASURES OF A GOOD RESULT HERE>
- Prior module output, if any: <INSERT PRIOR OUTPUT OR WRITE NONE HERE>
- Intended next module or human recipient: <INSERT NEXT STEP OR RECIPIENT HERE>

TASK
1. Check whether the supplied inputs are sufficient, relevant, authorised and internally consistent. List missing or conflicting items before continuing.
2. Plan and complete only the "explanation component" component. The explanation component performs the bounded explanation component function within the parent module and returns an inspectable intermediate result for Recommendation and Ranking.
3. Apply the module's core operation: score and order potential items against explicit goals, preferences, constraints, and diversity requirements.
4. Where useful, compare reasonable options, explain trade-offs, and identify dependencies. Related module connections recorded in the catalogue are: MOD-023 Keyword and Metadata Search; MOD-024 Semantic Retrieval; MOD-042 Output Quality Evaluation; MOD-020 Constraint Checking; MOD-044 Human Review and Approval; MOD-009 Draft Revision.
5. Preserve provenance by linking every material claim or transformed item to <INSERT SOURCE IDENTIFIER, PAGE, SECTION, TIMESTAMP OR RECORD ID HERE>.
6. Test the result against <INSERT QUALITY, SAFETY, PRIVACY OR ACCEPTANCE CRITERIA HERE>. Mark uncertainty and do not silently fill gaps.
7. Prepare a clear hand-off that can contribute to these expected module outputs: ranked options, reasons, and uncertainty.

OUTPUT FORMAT
Return the answer under exactly these headings:
1. Goal and scope
2. Input sufficiency check
3. Missing or conflicting information
4. explanation component result
5. Evidence and provenance map
6. Options, trade-offs and dependencies
7. Assumptions, uncertainties and risks
8. Hand-off package for <INSERT NEXT MODULE OR PERSON HERE>
9. Validation checklist
10. Human decision required

HUMAN REVIEW
A named person sets the acceptance threshold and verifies the result using these checks: criteria sensitivity test; diversity audit; human comparison. The accountable reviewer is <INSERT REVIEWER NAME OR ROLE HERE>. Do not treat the work as approved until that person records <INSERT APPROVAL, CORRECTION OR ESCALATION DECISION HERE>.

Local model prompt

GOAL
Complete one clearly bounded step for <INSERT PROJECT GOAL HERE>: the "explanation component" component of MOD-026 Recommendation and Ranking.

ROLE
Act as a careful information-retrieval specialist. Use only the text and data pasted into this prompt. Do not browse, call tools, invent missing information, or complete other module components.

WHAT TO INSERT
- Project or document name: <INSERT NAME HERE>
- Short project goal: <INSERT GOAL HERE>
- Audience: <INSERT AUDIENCE HERE>
- Source text or data: <INSERT SOURCE MATERIAL HERE>
- Required inputs (option set, selection criteria, and user history): <INSERT EACH INPUT HERE>
- Rules or constraints: <INSERT RULES, LIMITS OR REQUIRED FORMAT HERE>
- Required result: <INSERT WHAT THE USER NEEDS FROM THIS COMPONENT HERE>
- Source labels: <INSERT PAGE, SECTION, TIMESTAMP OR RECORD IDS HERE>

TASK
1. Restate the goal in one sentence.
2. List the inputs actually supplied. If anything needed is absent, write "MISSING INPUT" and name it. Do not guess.
3. Perform only this component task: This submodule implements the explanation step needed to score and order candidate items against explicit goals, preferences, constraints, and diversity requirements. It returns an inspectable intermediate result for Recommendation and Ranking.
4. Keep names, numbers, quotations and source labels exactly as supplied.
5. Check the result against <INSERT SIMPLE ACCEPTANCE CHECKS HERE>.
6. Prepare a short result that can support these expected module outputs: ranked options, reasons, and uncertainty.

OUTPUT FORMAT
Use exactly this structure:
GOAL: <one sentence>
INPUTS USED: <bulleted list>
MISSING INPUTS: <bulleted list or NONE>
EXPLANATION COMPONENT RESULT: <the completed bounded result>
SOURCE LINKS: <source label for each important item>
UNCERTAINTIES: <bulleted list or NONE>
HAND-OFF: <what the next person or module receives>
CHECK: <PASS, FAIL or NEEDS REVIEW, followed by one short reason>
HUMAN REVIEW: <INSERT REVIEWER NAME OR ROLE HERE must approve, correct or escalate>

STOP CONDITIONS
Stop and ask for more information if the source is missing, contradictory, unreadable, unauthorised, or too large for the available context. A named person sets the acceptance threshold and verifies the result using these checks: criteria sensitivity test; diversity audit; human comparison.