Contributing

ContentRX is private-taxonomy / public-evidence-of-work. The underlying rule set is not published; the model evolves under maintainer-only review. The public surfaces — accuracy reporting at /accuracy, the weekly /calibration log, and the named-expert essays — track how the model performs over time. See the private-taxonomy ADR for the rationale.

If you're a customer with feedback on a specific verdict, the most useful path is the in-product override — every dismissal is captured in the substrate and surfaces in the founder's review queue.

If you're a researcher or engineer interested in the methodology (situation-aware classification, content-type routing, severity thresholds, audience gating), open a discussion on the engine repository. The methodology is public; the rules are not.

Reporting a bug

Issues against the evaluator's output (false positives, false negatives, surface bugs in the Figma plugin / CLI / MCP server / LSP / GitHub Action) are welcome on the engine repo with the bug label. Please include:

• The string evaluated.
• The expected verdict and the verdict you actually got.
• The surface you saw it on (plugin / CLI / GitHub Action / etc.) and a recent timestamp so the team can correlate against the override stream.

Reports never include taxonomy IDs because the public surface doesn't render them; we'll match against the substrate on the maintainer side.