Skip to content

gz adr evaluate

Evaluate ADR and OBPI quality using deterministic scoring across 8 ADR dimensions and 5 OBPI dimensions. Produces an EVALUATION_SCORECARD.md in the ADR directory with a GO / CONDITIONAL GO / NO GO verdict.


Usage

Bash
gz adr evaluate <adr_id> [--json] [--no-scorecard]

What It Evaluates

ADR Quality (8 weighted dimensions)

# Dimension Weight
1 Problem Clarity 15%
2 Decision Justification 15%
3 Feature Checklist Completeness 15%
4 OBPI Decomposition Quality 15%
5 Lane Assignment Correctness 10%
6 Scope Discipline 10%
7 Evidence Requirements 10%
8 Architectural Alignment 10%

OBPI Quality (5 dimensions per brief)

Dimension Question
Independence Can this OBPI be completed without waiting for others?
Testability Can completion be verified with commands?
Value What concrete capability would be lost if removed?
Size Is this a 1-3 day work unit?
Clarity Could a different agent implement this without ambiguity?

Scaffold Detection

Briefs containing template placeholders (src/module/, First constraint, etc.) are flagged and scored low on the value dimension. This catches auto-generated stubs that were never authored.


Verdict Thresholds

ADR Weighted Total Verdict
>= 3.0 GO — Ready for proposal/defense review
2.5 - 3.0 CONDITIONAL GO — Address weaknesses, then re-evaluate
< 2.5 NO GO — Structural revision required

OBPI threshold: Average >= 3.0 per OBPI. Any dimension scoring 1 must be revised.


When to Use

  • After drafting a new ADR and its OBPIs — quality gate before proposal
  • Before moving a Draft ADR to Proposed / human defense review
  • When benchmarking the quality of an existing ADR package
  • After populating OBPI briefs — verify scaffold is cleared

Examples

Bash
# Evaluate a specific ADR
uv run gz adr evaluate ADR-0.3.0

# Machine-readable output
uv run gz adr evaluate ADR-0.3.0 --json

# Skip writing the scorecard file
uv run gz adr evaluate ADR-0.3.0 --no-scorecard

Options

Option Description
adr_id ADR identifier (e.g., ADR-0.3.0)
--json Emit machine-readable output
--no-scorecard Skip writing EVALUATION_SCORECARD.md to the ADR directory

Output

Writes EVALUATION_SCORECARD.md in the ADR package directory containing:

  • All ADR dimension scores with weighted totals
  • All OBPI dimension scores with averages
  • Overall verdict (GO / CONDITIONAL GO / NO GO)
  • Action items for any deficiencies

Pipeline Enforcement

When a scorecard exists with a NO GO verdict, gz obpi pipeline Stage 1 treats it as a blocker and aborts. This makes the evaluation a blocking gate for pipeline execution — run gz adr evaluate before starting OBPI work, and address NO GO action items before invoking the pipeline.

GO and CONDITIONAL GO verdicts do not block. Missing scorecards do not block (evaluation is optional until run).


Command Relationship
gz adr audit-check Post-implementation evidence verification (downstream)
gz obpi validate Single-brief or batch authored/completion readiness check
gz obpi validate --adr Batch-validate all briefs under an ADR
gz adr report Status view without scoring