site2voice 0.4.0

Generate agent-ready VOICE.md packs and CI benchmarks for website copy.

site2voice

Turn any website into agent-ready copy context. Then test whether the agent followed it.

pipx install site2voice

site2voice init https://example.com
site2voice bench https://example.com draft.md --strict

No LLM dependency. No browser dependency. Just a small Python CLI that extracts measurable voice signals and turns them into files coding agents can use.

Why

Agents are getting better at building interfaces, but they still guess the words: headings, CTAs, navigation labels, claim boundaries, and product tone.

DESIGN.md gives agents visual taste. site2voice gives them copy taste.

Voice Packs

The repo includes downloadable voice packs for well-known product and editorial sites: Apple, Stripe, Linear, Vercel, Notion, Figma, Shopify, GitHub, OpenAI, Anthropic, Hypebeast, Highsnobiety, Monocle, Wallpaper, and EYESMAG.

curl -L https://raw.githubusercontent.com/SihyeonJeon/site2voice/main/packs/stripe/VOICE.md -o VOICE.md

See packs.

What You Get

.site2voice/
  VOICE.md          human-readable writing brief
  voice.json        machine-readable voice profile
  agent-prompt.md   drop-in instruction for Codex, Claude Code, Cursor, etc.

The profile includes:

• heading shape, sentence rhythm, paragraph length, CTA shape;
• repeated vocabulary and CTA verbs;
• deterministic tone labels;
• unsupported-claim detection;
• copied-span detection.

Benchmark

site2voice bench scores candidate copy against a source voice profile and can fail CI when the copy drifts.

site2voice bench examples/editorial-home.html examples/after-copy.md --strict

Example fixture:

Candidate	Result	Overall	Lexicon	Copy safety
after-copy	PASS	83.8	70.0	93.2
before-copy	FAIL	36.6	0.0	100.0

The score rewards measurable voice alignment without rewarding verbatim copying.

Commands

# Generate one Markdown brief
site2voice https://example.com --out VOICE.md

# Generate an agent-ready context pack
site2voice init https://example.com --dir .site2voice

# Write JSON for custom pipelines
site2voice https://example.com --format json --out voice.json

# Gate candidate copy in CI
site2voice bench https://example.com draft.md \
  --fail-under 75 \
  --min-copy-safety 85 \
  --min-claim-safety 75

What It Is Not

• Not an official brand guideline.
• Not a DESIGN.md visual-token extractor.
• Not a crawler for private pages or authenticated apps.
• Not a tool for copying another site's prose.

Develop

python3 -m pip install -e .
make test
make bench-ci

Docs

• Benchmark
• Voice packs
• CI usage
• Agent workflows
• Voice patterns
• JSON schema
• Market position
• Research
• Source candidates
• Awesome eligibility
• Roadmap

License

MIT