What shipped
Built a focused standard-library Python CLI package, ai_pilot_proof_room, with three commands:
sample-proof-room: seeds a synthetic compatible workspace.generate-proof-room: creates buyer-facing proof-room artifacts fromsales-pack/sales-pack.jsonand optional control-tower context.validate-proof-room: enforces source contracts, artifact coverage, notices, scorecard structure, JSON parseability, and blocked unsafe autonomy language.
The generated proof room includes the demo runbook, 7-minute script, stakeholder proof pages, evaluation scorecard, approval-boundary walkthrough, objection handling, follow-up emails, and consolidated JSON export.
Architecture
- Used Python 3.11+ standard library only for easy local inspection and deterministic execution.
- Kept the project file-based: JSON in, Markdown/CSV/JSON out.
- Reused the parent build's contract style and safety language, but did not depend on the parent path at runtime.
- Made validation strict enough to catch missing files, removed notices, missing scorecard coverage, and unsafe autonomy promises.
- Kept sample data synthetic and embedded so verification works without network access or external credentials.
Trimmed scope
- No web UI, database, hosted deployment, PDF export, slide deck, or live integration.
- No external APIs, browser automation, OAuth, provider calls, or model calls.
- No attempt to regenerate the full parent sales pack. This build only needs the source JSON contract and optional control-tower context.
Limitations
- Generated materials are deterministic templates, not custom legal, compliance, safety, or production freight guidance.
- Validation uses explicit text coverage checks and unsafe phrase matching rather than semantic review.
- The optional control-tower contract is read for workflow context but not deeply validated beyond local safety and opportunity presence.
- The scorecard is fixed to five pilot criteria in the MVP.
Suggested next steps
- Add support for importing a real sanitized sales-pack JSON from a prior local build.
- Add a
--strict-control-towerflag for deeper optional upstream validation. - Add alternate stakeholder templates for shipper, broker, 3PL, and consulting-partner audiences.
- Add a local HTML preview after the Markdown/CSV workflow proves useful.