pendo-io/novus
Novus skills for coding agents: UX review and evidence-backed next-work and portfolio investment decisions, powered by Pendo product analytics.
Explain what a team's shipped improvements add up to for customers in a product area, why they matter, and what evidence supports their value. Also use to verify a specific change, launch, or experiment, or when a builder explicitly requests an individual impact report. Portfolio allocation belongs to build-investment.
Use when a leader asks whether engineering investment is going to the right goals or product areas, whether to sustain or change that investment, or needs a planning-cycle or portfolio review. Explaining the value of a team's shipped improvements belongs to build-impact.
Use when a builder wants current usage for the product area a PR, ticket, or proposed feature touches, and an evidence-based read of the customer benefit it could create. Can prepare a PR or ticket comment and post it after confirmation. Use build-impact to evaluate shipped outcomes.
Reviews uncommitted or branch-local code changes for real UX problems before a PR is opened — wrong visual semantics, unpersisted UI state, poor discoverability, missing loading and error states, removal of well-used paths, redundant UI. When the Novus MCP server is connected it backs findings with product analytics and applies the team's UX review instructions (PR Workflows settings and product-wiki preferences) so the local pass matches the pull request review; it falls back to code-observable findings when it is not. Use when the user asks for a UX review, a design review, a pre-PR check, whether their current changes will hurt users, or to save or update a UX review preference for the team.
Use when a decision depends on Pendo/Novus behavioral data and you need to know whether the measured surface is instrumented well enough to trust — before starting work, before reading impact numbers, or as the measurement-trust sub-check for build-impact, build-investment, whats-next, or usage-brief. It returns a trust verdict and the smallest exact repair; it never adds instrumentation.
Use when a builder asks what to build next, whether to finish, continue, narrow, pause, or switch current engineering work, what to defer, or when an autonomous coding agent must steer among bounded work options. Not for team-level portfolio allocation or designing an experiment for an already selected outcome.