Proposal: add skill-grader — a quality-rubric evaluation skill

**Summary**

I'd like to propose adding a new skill, `skill-grader`, which evaluates other Agent Skills against a production-grade quality rubric: 3 hard gates (YAML frontmatter, SKILL.md < 500 lines, description ≤ 1024 chars) + six-dimension weighted scoring (trigger design / workflow discipline / structure / output contract / guardrails / evaluation loop), producing evidence-based reports.

**Rationale**

This repo currently covers creation (`skill-creator`), document generation (`docx`/`pdf`/`pptx`/`xlsx`), and domain examples — but has no skill that *evaluates* skill quality. A grading skill completes the loop: create → evaluate → improve. It also operationalizes the guidance in the spec (progressive disclosure, optimized descriptions, evaluation).

**Implementation**

- Self-contained: SKILL.md (~180 lines) + examples/score_evals.md (3 self-test cases)
- No external dependencies; rubric embedded inline
- MIT licensed (can relicense to Apache-2.0 to match this repo if preferred)
- Reference implementation: https://github.com/yt2102/skill-grader

**Questions**

1. Does this repo accept community-contributed skills? (No CONTRIBUTING.md found.)
2. If yes, any requirements for submission (license, tests, spec compliance)?

Start by reviewing the reference implementation at https://github.com/yt2102/skill-grader and the repository's existing skill layout. Confirm how community skills, licensing, and self-tests are handled, then add SKILL.md and examples/score_evals.md with the proposed three self-test cases. Done means the skill is self-contained, follows the repository's accepted requirements, and its evaluation rubric is exercised by the examples.

Written by the indexing model from the issue text.