EVT 2026-02-21 - Stage B/D Goal Engine V1: Depth, Mapping, Versioning, Confidence Gate (Autocomplete)

Session scope

Continue autonomous completion of goal modeling so it remains accurate across programs while keeping UX lean and implementation scalable.

Confirmed outcomes (auto-recommended defaults)

1. Goal depth scope lock:

• goal config depth is program + assessment_form + skill (B-level depth),
• avoids global one-size score target.

2. Scale conversion guardrail:

• conversion/normalization is allowed only via explicit approved mapping table,
• no auto-convert-all-to-percent fallback by default.

3. Multi-goal structure:

• support one primary goal and optional secondary goal,
• primary drives default Home/Result emphasis and recommendation priority.

4. Goal versioning:

• each goal update creates immutable version,
• attempt/result comparison uses goal version effective at submission time.

5. Numeric goal-gap confidence gate:

• show numeric distance-to-goal only when there are at least 3 comparable attempts in latest 30 active days,
• otherwise keep trend/progress view only.

6. Fallback behavior when incompatible/low-confidence:

• no numeric goal-gap surface,
• recommendation bias switches to same assessment_form (and same skill where possible) to collect comparable evidence.

Open items carried forward

• None for baseline lock.