v1.4.1: guarantee on-brand WCAG AA + test all output formats + refres… #10

	name: Validate Skill

	# Runs whenever the skill definition or its engine changes, plus on PRs and
	# manual dispatch. Cheap (Python stdlib only, ~15s).
	on:
	push:
	branches: [main]
	paths:
	- 'SKILL.md'
	- 'INSTRUCTIONS.md'
	- 'scripts/**'
	- 'agents/**'
	- '.github/workflows/validate.yml'
	pull_request:
	branches: [main]
	workflow_dispatch:

	jobs:
	validate:
	runs-on: ubuntu-latest
	steps:
	- uses: actions/checkout@v4

	- name: Set up Python
	uses: actions/setup-python@v5
	with:
	python-version: '3.12'

	- name: Validate SKILL.md frontmatter
	# Equivalent to WorkBuddy's `package_skill.py` validation:
	# required fields, hyphen-case name, no angle brackets in description.
	run: python scripts/validate_skill.py .

	- name: Generator smoke + robustness + golden + contrast tests
	# Same script contributors run locally, so "passes CI" == "passes
	# scripts/smoke_test.py". Covers: cool/warm/balanced brands produce
	# non-empty tokens.css + preview.html; every --color-* value is valid
	# hex; semantic sanity (guards the HSL->RGB regression where a blue
	# brand emitted a blue "error"); 9 frozen golden token values for
	# #2563EB plus WCAG AA contrast assertions on text/surface, dark text/bg
	# and on-brand; extreme/degenerate brands (black/white/gray/short/
	# lowercase) don't crash; invalid input is rejected with a non-zero exit.
	run: python scripts/smoke_test.py

	- name: Summary
	if: always()
	run: echo "Validation complete — frontmatter + generator smoke + extreme-color robustness + semantic sanity."

Navigation Menu