What shipped
Built a standard-library Python CLI package named agency_ai_visibility_pilot_pack with three commands:
sample: creates a seeded local demo workspace foreast-nashville-dentist,brentwood-hvac, andgermantown-cafepilot-pack: readswatchboard/watchboard.jsonand writes the agencysales-pack/validate-pilot-pack: performs deterministic structural, pricing, demo coverage, and compliance checks
The build includes README.md, BUILD_SUMMARY.md, verify.sh, package source code, demo fixture generation, Markdown/CSV/JSON renderers, and validation logic.
Architecture
- Python 3.11+ standard library only for easy local inspection and no dependency installation requirement.
- Treat the parent Local AI Visibility Watchboard JSON shape as the stable contract.
- Generate deterministic local files and avoid hosted services, databases, browser automation, network calls, OAuth, API keys, and provider/model calls.
- Keep compliance language centralized in
schema.pyso Markdown, CSV, JSON, and validation all share the same boundaries.
Trimmed scope
Intentionally excluded:
- Web UI
- Scheduler
- Database
- CRM integration
- Email sending
- Billing
- Google Business Profile API
- Live search checks
- Scraping
- Browser automation
- OpenAI, Claude, Anthropic, OpenRouter, Gemini, Google, GitHub, or other provider/API calls
Limitations
- The demo watchboard is a local fixture, not live evidence.
- The generator relies on the supplied
watchboard.jsonquality and does not deeply normalize arbitrary parent schema drift. - Output language is template-based and should be edited by a human before sales use.
- Validation checks required phrases and structure, but it is not a legal or platform-policy review.
Suggested next steps
- Add a small importer for real parent watchboard workspaces with schema drift warnings.
- Add optional template override files for Michael's agency voice.
- Add snapshot tests for generated Markdown and CSV files.
- Add a red-team validator mode for overclaiming, review policy risk, and missing source-note evidence.