/claim #18

Adds a self-contained scientific-bounty-award-publication-guard package for the Scientific Bounty System.

This slice runs after an award decision and before the platform publishes winner/result metadata. It checks whether public award publication is safe with respect to payout authorization, appeal windows, IP/license readiness, embargo dates, private challenge redaction, named-winner consent, award split validity, recipient publication terms, and reproducibility evidence.

Why this is distinct:

• not intake/rubric scoring
• not escrow funding readiness
• not export-control screening
• not duplicate submission provenance
• not IP policy selection
• not appeal handling itself
• not payout routing

Artifacts included:

• deterministic sample packets
• JSON results
• Markdown reviewer report
• SVG summary
• short MP4 demo artifact

Validation run locally:

• npm test
• npm run demo
• node –check src/index.js
• node –check scripts/demo.js
• node –check test/awardPublicationGuard.test.js
• git diff –check

Boundaries:

• synthetic data only
• no payment processors
• no private submissions
• no credentials
• no external APIs
• no legal advice

AI-assisted with OpenAI Codex; I reviewed the implementation and kept the module isolated.