enhance(#1579): deterministic gsd-tools query eval.score verb (#1583)

* feat(#1579): deterministic gsd-tools query eval.score verb

Split C of #1573 (pure code, lowest risk). Adds an eval.score query verb
(coverage*0.6 + infra*0.4; bands 80/60/40) mirroring the verify.* chain;
gsd-eval-auditor consumes it instead of doing weighted arithmetic in-prompt.
Non-breaking — additive only.

arXiv: 2601.15130 (Plausibility Trap/DPDM), 2507.10281 (Table Agent), 2508.15754 (TIR).

* fix(#1579): address review — domain guard, property test, glossary, SKIP_ROOT, inventory/baseline

- C3 input-domain: reject out-of-domain eval.score (require 0<=covered<=total; was emitting overall_score>100 / negatives)
- C1 property test: add tests/eval.property.test.cjs (fast-check) — determinism, band monotonicity, [0,100] bounds, never-throws
- C2 glossary: CONTEXT.md "Eval Scoring Module" entry (source-of-truth path + interface)
- C4: add `eval` to SKIP_ROOT_RESOLUTION (pure arithmetic; no .planning/ access)
- inventory: register generated eval.cjs/eval-command-router.cjs (INVENTORY-MANIFEST.json + INVENTORY.md rows)
- size: regen agent-size baseline for gsd-eval-auditor (reused gsd_run shim + eval.score step)
- eslint: ignore generated eval*.cjs (ADR-457 bin/lib migration coverage)

* fix(#1579): register eval family in alias-drift gates

Add EVAL_COMMAND_ALIASES/EVAL_SUBCOMMANDS to scripts/check-alias-drift.cjs
families and to familyArrayKeys in the manifest-coverage test, so the eval
family lands under the same drift guard as every sibling family
(state/verify/init/phase/phases/validate/roadmap). Addresses trek-e review.

check:alias-drift ok; feat-3251 coverage 9/9; eval suites 10/10.

* docs(#1579): use half-open verdict band ranges in CLI-TOOLS

overall_score is fractional and thresholds are >=80/>=60/>=40, so a score
in [79,80) is correctly NEEDS WORK despite the old '60-79' label. Relabel
bands as 60-<80 / 40-<60 / 0-<40 to match the code. Addresses trek-e nit.

* fix(#1579): validate eval.score CLI inputs

Reject unknown infra tokens and fractional counts, and pin the 80-point verdict boundary including rounding-before-banding behavior.
This commit is contained in:
Alex V.
2026-06-25 00:16:13 +03:00
committed by GitHub
parent a63684c222
commit 1a46109b97
17 changed files with 409 additions and 11 deletions

View File

@@ -248,6 +248,42 @@ This command is strictly read-only — no config writes, no disk mutation.
---
### `query eval.score`
```bash
node gsd-tools.cjs query eval.score --covered <N> --total <N> --infra <tooling>,<dataset>,<cicd>,<guardrails>,<tracing>
```
Deterministic scorer for eval-auditor results. Computes coverage, infrastructure, and overall scores from audited inputs. Called by `gsd-eval-auditor` in its `calculate_scores` step — agents must not recompute these values by hand.
**Inputs:**
| Flag | Type | Description |
|---|---|---|
| `--covered` | integer | Number of eval dimensions scored COVERED |
| `--total` | integer | Total planned eval dimensions |
| `--infra` | string | Comma-separated list of 5 infra component statuses (order: tooling, dataset, cicd, guardrails, tracing); each value is `ok`, `partial`, or `missing` |
**Output JSON:**
| Field | Type | Description |
|---|---|---|
| `coverage_score` | number | `covered / total × 100` |
| `infra_score` | number | `(sum of component weights) / 5 × 100` (`ok`=1, `partial`=0.5, `missing`=0) |
| `overall_score` | number | `(coverage_score × 0.6) + (infra_score × 0.4)` |
| `verdict` | string | `PRODUCTION READY` (80–100) / `NEEDS WORK` (60–<80) / `SIGNIFICANT GAPS` (40–<60) / `NOT IMPLEMENTED` (0–<40) |
**Example:**
```bash
node gsd-tools.cjs query eval.score --covered 3 --total 5 --infra ok,partial,missing,ok,ok
# → {"coverage_score":60,"infra_score":70,"overall_score":64,"verdict":"NEEDS WORK"}
```
This command is strictly read-only — no config writes, no disk mutation.
---
## Model Resolution
```bash