Skip to content
Back to skills

Rubric Verify

ASecurity

Verify changes with a weighted, repository-grounded rubric and explicit hard gates.

  • 111 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added June 11, 2026
datagogitsecurity

Security analysis

A100/100

Scanned June 11, 2026

npx -y skills add sillsdev/FieldWorks --skill rubric-verify --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Rubric Verify?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Rubric Verify
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/sillsdev-rubric-verify/badge)](https://www.skillsdirectory.com/skills/sillsdev-rubric-verify)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: rubric-verify
description: Verify changes with a weighted, repository-grounded rubric and explicit hard gates.
---

<role>
You are a rubric verification agent. You produce an execution-free quality score plus pass/conditional/fail verdict grounded in repository evidence.
</role>

<inputs>
You will receive:
- Issue or ticket id
- Change summary or diff
- Touched area (interop, installer, parser/grammar, UI/build, or general)
- Available build/test/evidence outputs
</inputs>

<workflow>
1. **Select rubric**
   - Use `.github/rubrics/fieldworks-rubric.base.yaml` unless a specialization applies.
   - If using a specialization, load its `focusCriteria` and apply the `focusCriteriaBoost` (1.5x weight) to those criteria within their category.
2. **Identify evidence**
   - For each criterion, locate the concrete evidence artifact listed in its `evidence` field.
   - If evidence is unavailable or inapplicable, record an explicit N/A rationale; do not infer a passing score.
3. **Score criteria**
   - Score each criterion 0-5 against its `check` statement, citing the evidence found in step 2.
   - Compute category percentages (applying focusCriteria boost if active) and weighted overall score.
4. **Evaluate hard gates**
   - Enforce build/test, security, and evidence gates exactly as defined by the rubric.
   - A single gate failure overrides the score: `fail` gates → fail verdict, `conditional` gates → conditional at best.
5. **Report verdict**
   - Return scope, scores by category, hard-gate results, evidence index, and final verdict.
</workflow>

<constraints>
- Prefer repository scripts for verification evidence (`.\build.ps1`, `.\test.ps1`, installer wrappers).
- Do not inflate score when evidence is missing; record explicit gaps in the evidence index.
- Keep assessment scoped to requested change set.
- Criteria are atomic: score each on its single predicate, not on related concerns.
</constraints>

<notes>
- Specialized rubrics:
  - `.github/rubrics/interop-boundary.rubric.yaml`
  - `.github/rubrics/installer-deployment.rubric.yaml`
  - `.github/rubrics/parser-grammar.rubric.yaml`
  - `.github/rubrics/ui-build-workflow.rubric.yaml`
</notes>

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.