使い方
- インストールコマンドをコピーし、自分の環境へ導入する前にソースを確認します。
- 必要なツール、認証情報、対応環境を原文で確認します。
- 入力と成果物を指定し、出力を資料と照合します。
npx skills add https://github.com/yogsoth-ai/de-anthropocentric-research-engine --skill aggregate-ranking開始前の確認
- 導入にはモデル契約、API枠、ファイルへのアクセス許可は含まれません。
- スター数や導入数は活動の指標であり、独立した品質・安全性試験ではありません。
リポジトリの原文
見出し、例、コード、表、リンク、参照画像を含む原文を表示しています。
aggregate-ranking
Purpose
Aggregate criterion or comparison results into an ordered recommendation under an explicit rule.
Input contract
yaml
required: [candidate_set, criterion_results, aggregation_rule]
optional: [tie_break_rule, missing_value_policy, uncertainty_annotations]
constraints: [criterion directions and scales must be declared; no silent imputation]Procedure
- Align candidate identifiers, criterion directions, units, and validity flags.
- Apply the supplied aggregation rule without changing weights or directions.
- Propagate missingness and uncertainty; apply the declared tie-break only after aggregation.
- Return ordered candidates with component contributions and recommendation status.
Output contract
yaml
produces: [ordered_recommendation, aggregate_scores, contribution_table, unresolved_comparisons]
delta_fields: [findings, evidence_updates, decisions, uncertainties]Quality gates
- Every ranked candidate has a traceable value for each required criterion or an explicit unresolved marker.
- Aggregation reproduces the supplied rule and preserves criterion direction.
- Ties and sensitivity to tie-breaks are reported.
Parameterization
Caller supplies candidate schema, criterion scales/directions, weights or aggregation formula, tie-break rule, and missing/uncertainty policy.
Failure and counterexamples
Reject mixed units without normalization; reject a recommendation when a hard criterion is unresolved.
Provenance map
- resolved: priority-synthesis
- resolved: scoring-synthesis
Preserved source criteria ledger
| source | criterion |
|---|---|
| priority-synthesis | All scoring dimensions are present for every gap. |
| priority-synthesis | Weight vector sums to 1.0 within +/-0.001. |
| priority-synthesis | Priority list is sorted descending; ties use feasibility sub-score. |
| priority-synthesis | Top N is N=min(3,total gaps) and includes attack-path suggestions. |
| scoring-synthesis | Final recommendation includes recommended alternative, confidence, key assumptions, and risk warnings. |
