Imported from yogsoth-ai/de-anthropocentric-research-engine (skills/evaluation-filtering/SKILL.md). Install upstream with npx skills add yogsoth-ai/de-anthropocentric-research-engine --skill evaluation-filtering. Copyright stays with the author.

Evaluation Filtering

Multi-dimensional evaluation and tiered filtering of generated ideas.

Stages

Stage 1: Novelty Assessment

Score all ideas on novelty dimensions using novelty-scoring SOP. Assign tier: BREAKTHROUGH / HIGH / MODERATE / INCREMENTAL.

Stage 2: Feasibility Signal

For HIGH and BREAKTHROUGH tier ideas, assess initial feasibility:

• Technical feasibility (can it be built with known methods?)
• Resource feasibility (reasonable cost/time/expertise?)
• Constraint satisfaction (does it meet stated requirements?)

This is a signal, not a deep validation — that's the convergence repo's job.

Stage 3: Ranking

Rank ideas by composite score: novelty (0.6) × feasibility signal (0.2) × completeness (0.2).

Stage 4: Selection

Select top ideas for output. All BREAKTHROUGH ideas pass regardless of feasibility. HIGH ideas pass if feasibility ≥ MEDIUM. MODERATE ideas pass only if they fill a coverage gap.

Minimum Yield

Available SOPs