rank-candidates
Purpose
Rank, classify, screen, or select typed candidates under explicit criteria and hard constraints.
Input contract
mode_contracts:
gap-prioritization: &rank_input
required: [candidates, criteria, decision_rule]
optional: [weights, evidence, hard_constraints, object_schema]
constraints: [criterion_direction_and_missing_value_policy_must_be_explicit]
direction-selection: *rank_input
mcda-best-choice: *rank_input
full-ranking: *rank_input
category-sorting: *rank_input
non-compensatory-screening: *rank_input
rapid-triage: *rank_input
stakeholder-weighted: *rank_input
Execution protocol
Do not perform called SOP operations inline; each loaded SOP owns its contract and thresholds.
- Normalize the candidate schema and separate hard constraints from preferences. You MUST load skill
define-criteria to derive explicit criteria and criterion directions from the objective and candidate set.
If the candidates cannot be made comparable because the research goal still contains unresolved branches, consider decompose-research-goal before ranking.
- Select
gap-prioritization, direction-selection, mcda-best-choice, full-ranking, category-sorting, non-compensatory-screening, rapid-triage, or stakeholder-weighted.
If candidates are better compared pairwise than scored absolutely, consider pairwise-ranking. If the decision requires selecting a jointly feasible subset rather than ordering independent candidates, portfolio-optimization may be the better next tactic. If feasibility is the binding uncertainty rather than relative merit, consider analyze-constraints-readiness.
- You MUST load skill
normalize-comparison-scale to put heterogeneous criterion values on a declared comparable scale. You MUST load skill score-object to score every typed candidate against the supplied rubric and evidence. You MUST load skill aggregate-ranking to combine the criterion or comparison outputs under the declared rule.
- You MUST load skill
assess-sensitivity to perturb the declared weights or decision inputs, report stability, and return ordered or categorized candidates with rationale.
If a prioritized gap or direction is ready to become a testable proposition, consider formulate-hypotheses as the next tactic.
Mode branches
For mcda-best-choice, full-ranking, category-sorting, or stakeholder-weighted, You MUST load skill elicit-weights to produce and validate the criterion-weight vector. For category-sorting or non-compensatory-screening, You MUST load skill set-threshold to justify the decision boundaries.
gap-prioritization: rank research gaps by evidence deficit and expected value so scarce investigation effort reaches the most consequential unknowns first. You MUST load skill normalize-gap to normalize heterogeneous gap records before ranking.
direction-selection: narrow competing research directions against explicit scope, evidence coverage, and feasibility constraints before committing to one. You MUST load skill assess-goal-feasibility to test candidate directions against resources, obstacles, and timeline.
mcda-best-choice: combine normalized multi-criteria scores with declared weights to select the strongest feasible option while preserving criterion-level rationale.
full-ranking: produce a complete ordered list using at least the required comparison methods, exposing incomparable pairs and method disagreement.
category-sorting: assign candidates to threshold-defined classes when ordinal categories are more defensible than fine-grained ranks.
non-compensatory-screening: apply hard thresholds and vetoes so a severe failure on one criterion cannot be hidden by strengths elsewhere. You MUST load skill apply-veto-filter to enforce hard failures. You MUST load skill check-dominance to expose dominated and non-dominated alternatives.
rapid-triage: use coarse importance and feasibility passes to reduce a large candidate set quickly, retaining elimination reasons for later review.
stakeholder-weighted: aggregate perspective-specific scores with an explicit consensus rule, showing where stakeholder rankings converge or diverge.
Output contract
mode_contracts:
gap-prioritization:
produces: [ranking_or_categories, scores, sensitivity_results, recommendation]
delta_fields: [findings, decisions, uncertainties, recommended_jumps]
direction-selection: &screening_output
produces: [ranking_or_categories, scores, eliminated_candidates, recommendation]
delta_fields: [findings, decisions, uncertainties, recommended_jumps]
mcda-best-choice:
produces: [scores, weights, sensitivity_results, recommendation]
delta_fields: [findings, decisions, uncertainties, recommended_jumps]
full-ranking: &weighted_ranking_output
produces: [ranking_or_categories, scores, weights, sensitivity_results, recommendation]
delta_fields: [findings, decisions, uncertainties, recommended_jumps]
category-sorting:
produces: [ranking_or_categories, scores, weights, recommendation]
delta_fields: [findings, decisions, uncertainties, recommended_jumps]
non-compensatory-screening: *screening_output
rapid-triage: *screening_output
stakeholder-weighted: *weighted_ranking_output
Thresholds and quality gates
- Full ranking: select >=2 ranking methods when the source protocol calls for method comparison.
- Weight elicitation: select >=2 weighting methods where required.
- Priority sensitivity: perturbation scenarios cover the declared weight space at a justified relative floor, each scenario annotated; retain the observed ranking-stability verdict.
- Direction narrowing imports report candidate/evidence-pool coverage, full-text coverage, independent-source ratio, batch increment, stopping reason, and source references rather than fixed paper/page counts.
- For every relative gate, declare the eligible candidate/evidence universe (denominator) and record the covered candidates or evidence items (numerator), the batch increment, the stopping reason, and source references; a gate is not passable when any audit field is missing.
- Stop direction search when marginal information gain falls below the justified floor across batches; report the resulting saturation state alongside coverage ratio and independent-source ratio.
- Non-compensatory and category modes must expose threshold/veto values; never hide them in prose.
Failure and counterexamples
Reject rankings with undefined criterion direction, unhandled missing values, or hard constraints treated as compensable scores. Do not collapse pairwise active ranking into scalar ranking.
Provenance map
20 architecture old entries; shared criteria->core->aggregate kernel retained. Portfolio and pairwise operations remain separate nodes, not silently absorbed.
Legacy context checkpoint / Delta notes
Append candidate set hash, criteria/weights, rule, ranking, sensitivity scenarios, exclusions, and unresolved trade-offs.
Preserved source criteria ledger
| source |
source line |
kind |
source criterion |
| multi-criteria-ranking |
49 |
numeric |
Each dimension is scored independently (1-5) to avoid cross-contamination between dimensions. Weights are set by AHP (Analytic Hierarchy Process) or specified by the user. Final score = $\Sigma$(dimension score * dimension weight). |
| multi-criteria-ranking |
51 |
numeric |
Sensitivity check: perturb weights by +/-20%; if the ranking is unchanged the conclusion is robust; if the ranking flips it must be flagged as "weight-sensitive". |
| multi-criteria-ranking |
53 |
textual |
## Budget Gate |
| multi-criteria-ranking |
57 |
numeric |
\ |
| multi-criteria-ranking |
58 |
numeric |
\ |
| multi-criteria-ranking |
59 |
numeric |
\ |
| evidence-based-prioritization |
51 |
textual |
## Budget Gate |
| evidence-based-prioritization |
55 |
numeric |
\ |
| evidence-based-prioritization |
56 |
numeric |
\ |
| evidence-based-prioritization |
57 |
numeric |
\ |
| stakeholder-weighted-ranking |
35 |
textual |
- A consensus must be built across parties, or the ranking differences across perspectives must be shown |
| stakeholder-weighted-ranking |
58 |
textual |
## Budget Gate |
| stakeholder-weighted-ranking |
62 |
numeric |
\ |
| stakeholder-weighted-ranking |
63 |
numeric |
\ |
| stakeholder-weighted-ranking |
64 |
numeric-table |
\ |
| rapid-triage |
49 |
numeric |
Round 2: light scoring (1-3 points, two dimensions) |
| rapid-triage |
51 |
numeric |
- Importance (1-3): a rough estimate of field impact |
| rapid-triage |
52 |
numeric |
- Feasibility (1-3): whether progress can be made within 6 months with existing resources |
| rapid-triage |
56 |
textual |
Key insight: the three round-1 questions must be answered quickly (no more than 30 seconds per gap); no deep analysis allowed. Speed is the core value of this strategy. |
| rapid-triage |
58 |
textual |
## Budget Gate |
| rapid-triage |
60 |
numeric-table |
\ |
| rapid-triage |
62 |
numeric |
\ |
| rapid-triage |
63 |
numeric |
\ |
| rapid-triage |
64 |
numeric |
\ |
| rapid-triage |
71 |
numeric |
4. Call the importance-scoring SOP on the Keep set (1-3 coarse score) |
| rapid-triage |
72 |
numeric |
5. Call the feasibility-scoring SOP on the Keep set (1-3 coarse score) |
| priority-sensitivity-testing |
28 |
numeric |
This tactic first establishes baseline weights (AHP or equal weights), then systematically perturbs the weights (+/-20%), observes the ranking changes, and finally gives a stability verdict. |
| priority-sensitivity-testing |
35 |
numeric |
\ |
| priority-sensitivity-testing |
42 |
numeric |
2. weight-perturbation: apply +20% and -20% perturbations to each dimension in turn (the remaining dimensions are adjusted proportionally to keep the sum at 1), producing a ranking for each perturbation scenario |
| priority-sensitivity-testing |
47 |
numeric |
- Perturb only the highest-weight dimension (+/-20%), producing 2 perturbation scenarios |
| priority-sensitivity-testing |
52 |
numeric |
- weight-perturbation expands the perturbation range to +/-30% and adds extreme scenarios (one dimension's weight set to 0) |
| priority-sensitivity-testing |
55 |
textual |
## Minimum Yield |
| priority-sensitivity-testing |
58 |
numeric |
- Ranking results for at least 3 perturbation scenarios (each scenario annotated with its perturbation content) |
| priority-sensitivity-testing |
60 |
numeric |
- A final stability verdict: Stable (top N unchanged across all scenarios) / Partially Sensitive (1-2 changes in the top N) / Highly Sensitive (more than 2 changes in the top N) |
| best-option-selection |
17 |
numeric |
- Moderate number of candidates (3-15) |
| best-option-selection |
21 |
numeric |
\ |
| best-option-selection |
23 |
numeric |
\ |
| best-option-selection |
24 |
numeric-table |
\ |
| best-option-selection |
25 |
numeric-table |
\ |
| best-option-selection |
26 |
numeric-table |
\ |
| best-option-selection |
27 |
numeric-table |
\ |
| full-ranking |
22 |
numeric |
\ |
| full-ranking |
24 |
numeric |
\ |
| full-ranking |
25 |
numeric-table |
\ |
| full-ranking |
26 |
numeric-table |
\ |
| full-ranking |
27 |
numeric-table |
\ |
| full-ranking |
28 |
numeric-table |
\ |
| full-ranking |
29 |
numeric-table |
\ |
| full-ranking |
64 |
numeric |
2. Select >=2 ranking methods (recommended: PROMETHEE II + MAVT) |
| full-ranking |
66 |
textual |
4. Pay attention to incomparable pairs in partial orders (specific to ELECTRE III) |
| category-sorting |
22 |
numeric |
\ |
| category-sorting |
24 |
numeric |
\ |
| category-sorting |
25 |
numeric-table |
\ |
| category-sorting |
26 |
numeric-table |
\ |
| category-sorting |
27 |
numeric-table |
\ |
| category-sorting |
28 |
numeric-table |
\ |
| non-compensatory-screening |
27 |
numeric |
\ |
| non-compensatory-screening |
29 |
numeric |
\ |
| non-compensatory-screening |
30 |
numeric-table |
\ |
| non-compensatory-screening |
31 |
numeric-table |
\ |
| non-compensatory-screening |
32 |
numeric-table |
\ |
| non-compensatory-screening |
65 |
textual |
2. Invoke threshold-setting to define minimum thresholds for each criterion |
| non-compensatory-screening |
75 |
textual |
Screening Rule: Conjunctive rule (all criteria must be met) |
| weight-elicitation |
23 |
numeric |
\ |
| weight-elicitation |
25 |
numeric |
\ |
| weight-elicitation |
26 |
numeric |
\ |
| weight-elicitation |
27 |
numeric-table |
\ |
| weight-elicitation |
60 |
numeric |
2. Select >=2 weighting methods (recommended: AHP + BWM or Swing + Simos) |
| weight-elicitation |
76 |
numeric |
- AHP CR: [value] (< 0.1 pass) |
| direction-narrowing |
36 |
numeric |
- broad-paper-search: at least 80 papers scanned |
| direction-narrowing |
37 |
numeric |
- deep-web-search: at least 30 web pages read in full |
| present-and-ask |
29 |
numeric |
User's selected 1-2 fields of interest + reasoning. |
Context checkpoint / Delta notes
Append candidate set hash, criteria/weights, rule, ranking, sensitivity scenarios, exclusions, and unresolved trade-offs.
1---2name: rank-candidates3description: "Rank, classify, screen, or select typed candidates under explicit criteria and supplied hard constraints. Object schema and decision mode are parameters; pairwise and portfolio cases may jump to dedicated tactics."4metadata:5 source: https://skillmd.com/skills/yogsoth-ai/rank-candidates6---78# rank-candidates910## Purpose1112Rank, classify, screen, or select typed candidates under explicit criteria and hard constraints.1314## Input contract1516```yaml17mode_contracts:18 gap-prioritization: &rank_input19 required: [candidates, criteria, decision_rule]20 optional: [weights, evidence, hard_constraints, object_schema]21 constraints: [criterion_direction_and_missing_value_policy_must_be_explicit]22 direction-selection: *rank_input23 mcda-best-choice: *rank_input24 full-ranking: *rank_input25 category-sorting: *rank_input26 non-compensatory-screening: *rank_input27 rapid-triage: *rank_input28 stakeholder-weighted: *rank_input29```3031## Execution protocol3233Do not perform called SOP operations inline; each loaded SOP owns its contract and thresholds.34351. Normalize the candidate schema and separate hard constraints from preferences. You MUST load skill `define-criteria` to derive explicit criteria and criterion directions from the objective and candidate set.36 If the candidates cannot be made comparable because the research goal still contains unresolved branches, consider `decompose-research-goal` before ranking.372. Select `gap-prioritization`, `direction-selection`, `mcda-best-choice`, `full-ranking`, `category-sorting`, `non-compensatory-screening`, `rapid-triage`, or `stakeholder-weighted`.38 If candidates are better compared pairwise than scored absolutely, consider `pairwise-ranking`. If the decision requires selecting a jointly feasible subset rather than ordering independent candidates, `portfolio-optimization` may be the better next tactic. If feasibility is the binding uncertainty rather than relative merit, consider `analyze-constraints-readiness`.393. You MUST load skill `normalize-comparison-scale` to put heterogeneous criterion values on a declared comparable scale. You MUST load skill `score-object` to score every typed candidate against the supplied rubric and evidence. You MUST load skill `aggregate-ranking` to combine the criterion or comparison outputs under the declared rule.404. You MUST load skill `assess-sensitivity` to perturb the declared weights or decision inputs, report stability, and return ordered or categorized candidates with rationale.41 If a prioritized gap or direction is ready to become a testable proposition, consider `formulate-hypotheses` as the next tactic.4243## Mode branches4445For `mcda-best-choice`, `full-ranking`, `category-sorting`, or `stakeholder-weighted`, You MUST load skill `elicit-weights` to produce and validate the criterion-weight vector. For `category-sorting` or `non-compensatory-screening`, You MUST load skill `set-threshold` to justify the decision boundaries.4647- `gap-prioritization`: rank research gaps by evidence deficit and expected value so scarce investigation effort reaches the most consequential unknowns first. You MUST load skill `normalize-gap` to normalize heterogeneous gap records before ranking.48- `direction-selection`: narrow competing research directions against explicit scope, evidence coverage, and feasibility constraints before committing to one. You MUST load skill `assess-goal-feasibility` to test candidate directions against resources, obstacles, and timeline.49- `mcda-best-choice`: combine normalized multi-criteria scores with declared weights to select the strongest feasible option while preserving criterion-level rationale.50- `full-ranking`: produce a complete ordered list using at least the required comparison methods, exposing incomparable pairs and method disagreement.51- `category-sorting`: assign candidates to threshold-defined classes when ordinal categories are more defensible than fine-grained ranks.52- `non-compensatory-screening`: apply hard thresholds and vetoes so a severe failure on one criterion cannot be hidden by strengths elsewhere. You MUST load skill `apply-veto-filter` to enforce hard failures. You MUST load skill `check-dominance` to expose dominated and non-dominated alternatives.53- `rapid-triage`: use coarse importance and feasibility passes to reduce a large candidate set quickly, retaining elimination reasons for later review.54- `stakeholder-weighted`: aggregate perspective-specific scores with an explicit consensus rule, showing where stakeholder rankings converge or diverge.5556## Output contract5758```yaml59mode_contracts:60 gap-prioritization:61 produces: [ranking_or_categories, scores, sensitivity_results, recommendation]62 delta_fields: [findings, decisions, uncertainties, recommended_jumps]63 direction-selection: &screening_output64 produces: [ranking_or_categories, scores, eliminated_candidates, recommendation]65 delta_fields: [findings, decisions, uncertainties, recommended_jumps]66 mcda-best-choice:67 produces: [scores, weights, sensitivity_results, recommendation]68 delta_fields: [findings, decisions, uncertainties, recommended_jumps]69 full-ranking: &weighted_ranking_output70 produces: [ranking_or_categories, scores, weights, sensitivity_results, recommendation]71 delta_fields: [findings, decisions, uncertainties, recommended_jumps]72 category-sorting:73 produces: [ranking_or_categories, scores, weights, recommendation]74 delta_fields: [findings, decisions, uncertainties, recommended_jumps]75 non-compensatory-screening: *screening_output76 rapid-triage: *screening_output77 stakeholder-weighted: *weighted_ranking_output78```7980## Thresholds and quality gates8182- Full ranking: select >=2 ranking methods when the source protocol calls for method comparison.83- Weight elicitation: select >=2 weighting methods where required.84- Priority sensitivity: perturbation scenarios cover the declared weight space at a justified relative floor, each scenario annotated; retain the observed ranking-stability verdict.85- Direction narrowing imports report candidate/evidence-pool coverage, full-text coverage, independent-source ratio, batch increment, stopping reason, and source references rather than fixed paper/page counts.86- For every relative gate, declare the eligible candidate/evidence universe (denominator) and record the covered candidates or evidence items (numerator), the batch increment, the stopping reason, and source references; a gate is not passable when any audit field is missing.87- Stop direction search when marginal information gain falls below the justified floor across batches; report the resulting saturation state alongside coverage ratio and independent-source ratio.88- Non-compensatory and category modes must expose threshold/veto values; never hide them in prose.8990## Failure and counterexamples9192Reject rankings with undefined criterion direction, unhandled missing values, or hard constraints treated as compensable scores. Do not collapse pairwise active ranking into scalar ranking.9394## Provenance map959620 architecture `old` entries; shared criteria->core->aggregate kernel retained. Portfolio and pairwise operations remain separate nodes, not silently absorbed.9798## Legacy context checkpoint / Delta notes99100Append candidate set hash, criteria/weights, rule, ranking, sensitivity scenarios, exclusions, and unresolved trade-offs.101102## Preserved source criteria ledger103104| source | source line | kind | source criterion |105|---|---:|---|---|106| multi-criteria-ranking | 49 | numeric | Each dimension is scored independently (1-5) to avoid cross-contamination between dimensions. Weights are set by AHP (Analytic Hierarchy Process) or specified by the user. Final score = $\Sigma$(dimension score * dimension weight). |107| multi-criteria-ranking | 51 | numeric | **Sensitivity check**: perturb weights by +/-20%; if the ranking is unchanged the conclusion is robust; if the ranking flips it must be flagged as "weight-sensitive". |108| multi-criteria-ranking | 53 | textual | ## Budget Gate |109| multi-criteria-ranking | 57 | numeric | \\| S \\| 5-8 \\| >=3 dimensions \\| Optional \\| Ranking table + attack suggestions for top 2 gaps \\| |110| multi-criteria-ranking | 58 | numeric | \\| M \\| 9-15 \\| >=4 dimensions \\| Required \\| Ranking table + attack suggestions for top 3 gaps \\| |111| multi-criteria-ranking | 59 | numeric | \\| L \\| 16-20 \\| >=5 dimensions \\| Required (multi-weight scenarios) \\| Ranking table + attack suggestions for top 5 gaps + weight-sensitivity report \\| |112| evidence-based-prioritization | 51 | textual | ## Budget Gate |113| evidence-based-prioritization | 55 | numeric | \\| S \\| 3-8 \\| all 6 dimensions \\| >=2 supporting references per gap \\| ranking table + evidence-void report \\| |114| evidence-based-prioritization | 56 | numeric | \\| M \\| 9-15 \\| all 6 dimensions \\| >=3 supporting references per gap \\| ranking table + evidence-void report + attack suggestions for top 3 gaps \\| |115| evidence-based-prioritization | 57 | numeric | \\| L \\| 16-20 \\| all 6 dimensions \\| >=5 supporting references per gap \\| ranking table + detailed evidence map + attack suggestions for top 5 gaps \\| |116| stakeholder-weighted-ranking | 35 | textual | - A consensus must be built across parties, or the ranking differences across perspectives must be shown |117| stakeholder-weighted-ranking | 58 | textual | ## Budget Gate |118| stakeholder-weighted-ranking | 62 | numeric | \\| S \\| 5-10 \\| 2-3 classes \\| Simple average \\| Per-perspective rankings + consensus top-3 \\| |119| stakeholder-weighted-ranking | 63 | numeric | \\| M \\| 11-20 \\| 3-5 classes \\| Borda count \\| Per-perspective rankings + consensus top-5 + divergence analysis \\| |120| stakeholder-weighted-ranking | 64 | numeric-table | \\| L \\| 20+ \\| 5+ classes \\| Weighted Borda + sensitivity \\| Full perspective matrix + consensus ranking + divergence heatmap \\| |121| rapid-triage | 49 | numeric | **Round 2: light scoring (1-3 points, two dimensions)** |122| rapid-triage | 51 | numeric | - Importance (1-3): a rough estimate of field impact |123| rapid-triage | 52 | numeric | - Feasibility (1-3): whether progress can be made within 6 months with existing resources |124| rapid-triage | 56 | textual | **Key insight**: the three round-1 questions must be answered quickly (no more than 30 seconds per gap); no deep analysis allowed. Speed is the core value of this strategy. |125| rapid-triage | 58 | textual | ## Budget Gate |126| rapid-triage | 60 | numeric-table | \\| Tier \\| Input gap count \\| Round-1 retention rate \\| Round-2 output \\| Final output \\| |127| rapid-triage | 62 | numeric | \\| S \\| 50-80 \\| <=60% \\| top-15 \\| Candidate set + elimination-rationale summary \\| |128| rapid-triage | 63 | numeric | \\| M \\| 81-150 \\| <=50% \\| top-20 \\| Candidate set + elimination-rationale summary \\| |129| rapid-triage | 64 | numeric | \\| L \\| 150+ \\| <=40% \\| top-30 \\| Candidate set + elimination-rationale summary + category statistics \\| |130| rapid-triage | 71 | numeric | 4. Call the `importance-scoring` SOP on the Keep set (1-3 coarse score) |131| rapid-triage | 72 | numeric | 5. Call the `feasibility-scoring` SOP on the Keep set (1-3 coarse score) |132| priority-sensitivity-testing | 28 | numeric | This tactic first establishes baseline weights (AHP or equal weights), then systematically perturbs the weights (+/-20%), observes the ranking changes, and finally gives a stability verdict. |133| priority-sensitivity-testing | 35 | numeric | \\| weight-perturbation \\| Apply +/-20% perturbations to each dimension weight and recompute the ranking \\| Second step, systematic perturbation \\| |134| priority-sensitivity-testing | 42 | numeric | 2. weight-perturbation: apply +20% and -20% perturbations to each dimension in turn (the remaining dimensions are adjusted proportionally to keep the sum at 1), producing a ranking for each perturbation scenario |135| priority-sensitivity-testing | 47 | numeric | - Perturb only the highest-weight dimension (+/-20%), producing 2 perturbation scenarios |136| priority-sensitivity-testing | 52 | numeric | - weight-perturbation expands the perturbation range to +/-30% and adds extreme scenarios (one dimension's weight set to 0) |137| priority-sensitivity-testing | 55 | textual | ## Minimum Yield |138| priority-sensitivity-testing | 58 | numeric | - Ranking results for at least 3 perturbation scenarios (each scenario annotated with its perturbation content) |139| priority-sensitivity-testing | 60 | numeric | - A final stability verdict: **Stable** (top N unchanged across all scenarios) / **Partially Sensitive** (1-2 changes in the top N) / **Highly Sensitive** (more than 2 changes in the top N) |140| best-option-selection | 17 | numeric | - Moderate number of candidates (3-15) |141| best-option-selection | 21 | numeric | \\| Base SOP \\| Target \\| +/-10% Range \\| |142| best-option-selection | 23 | numeric | \\| criterion-definition \\| 5-8 criteria \\| 4-9 \\| |143| best-option-selection | 24 | numeric-table | \\| weight-elicitation-sop \\| 1 weight vector \\| 1 \\| |144| best-option-selection | 25 | numeric-table | \\| alternative-scoring \\| 1 score matrix \\| 1 \\| |145| best-option-selection | 26 | numeric-table | \\| normalization \\| 1 normalized matrix \\| 1 \\| |146| best-option-selection | 27 | numeric-table | \\| scoring-synthesis \\| 1 recommendation \\| 1 \\| |147| full-ranking | 22 | numeric | \\| Base SOP \\| Target \\| +/-10% Range \\| |148| full-ranking | 24 | numeric | \\| criterion-definition \\| 5-8 criteria \\| 4-9 \\| |149| full-ranking | 25 | numeric-table | \\| weight-elicitation-sop \\| 1 weight vector \\| 1 \\| |150| full-ranking | 26 | numeric-table | \\| alternative-scoring \\| 1 score matrix \\| 1 \\| |151| full-ranking | 27 | numeric-table | \\| normalization \\| 1 normalized matrix \\| 1 \\| |152| full-ranking | 28 | numeric-table | \\| rank-comparison \\| 1 agreement matrix \\| 1 \\| |153| full-ranking | 29 | numeric-table | \\| scoring-synthesis \\| 1 full ranking \\| 1 \\| |154| full-ranking | 64 | numeric | 2. Select >=2 ranking methods (recommended: PROMETHEE II + MAVT) |155| full-ranking | 66 | textual | 4. Pay attention to incomparable pairs in partial orders (specific to ELECTRE III) |156| category-sorting | 22 | numeric | \\| Base SOP \\| Target \\| +/-10% Range \\| |157| category-sorting | 24 | numeric | \\| criterion-definition \\| 5-8 criteria \\| 4-9 \\| |158| category-sorting | 25 | numeric-table | \\| weight-elicitation-sop \\| 1 weight vector \\| 1 \\| |159| category-sorting | 26 | numeric-table | \\| threshold-setting \\| 1 threshold set \\| 1 \\| |160| category-sorting | 27 | numeric-table | \\| alternative-scoring \\| 1 score matrix \\| 1 \\| |161| category-sorting | 28 | numeric-table | \\| scoring-synthesis \\| 1 classification \\| 1 \\| |162| non-compensatory-screening | 27 | numeric | \\| Base SOP \\| Target \\| +/-10% Range \\| |163| non-compensatory-screening | 29 | numeric | \\| criterion-definition \\| 3-5 screening criteria \\| 2-6 \\| |164| non-compensatory-screening | 30 | numeric-table | \\| threshold-setting \\| 1 threshold set \\| 1 \\| |165| non-compensatory-screening | 31 | numeric-table | \\| conjunctive-filter \\| 1 pass/fail list \\| 1 \\| |166| non-compensatory-screening | 32 | numeric-table | \\| dominance-check \\| 1 dominance report \\| 1 \\| |167| non-compensatory-screening | 65 | textual | 2. Invoke threshold-setting to define minimum thresholds for each criterion |168| non-compensatory-screening | 75 | textual | **Screening Rule:** Conjunctive rule (all criteria must be met) |169| weight-elicitation | 23 | numeric | \\| Base SOP \\| Target \\| +/-10% Range \\| |170| weight-elicitation | 25 | numeric | \\| criterion-definition \\| 5-8 criteria \\| 4-9 \\| |171| weight-elicitation | 26 | numeric | \\| weight-elicitation-sop \\| >=2 methods \\| 2-3 \\| |172| weight-elicitation | 27 | numeric-table | \\| rank-comparison \\| 1 comparison \\| 1 \\| |173| weight-elicitation | 60 | numeric | 2. Select >=2 weighting methods (recommended: AHP + BWM or Swing + Simos) |174| weight-elicitation | 76 | numeric | - AHP CR: [value] (< 0.1 pass) |175| direction-narrowing | 36 | numeric | - `broad-paper-search`: at least 80 papers scanned |176| direction-narrowing | 37 | numeric | - `deep-web-search`: at least 30 web pages read in full |177| present-and-ask | 29 | numeric | User's selected 1-2 fields of interest + reasoning. |178179## Context checkpoint / Delta notes180181Append candidate set hash, criteria/weights, rule, ranking, sensitivity scenarios, exclusions, and unresolved trade-offs.