Experiment: Agent-curated developer exploration trails

[sajeetharan]

Problem

Users without a precise query have little guided way to discover the ecosystem beyond rankings and activity streams.

Proposed Solution

Create evidence-backed, branching exploration trails assembled from real profiles around a transparent theme.

MVP Scope

• Curated seed themes
• Five to eight profiles per trail
• Each step explains why the profile belongs and cites public evidence
• Branch choices such as adjacent skill, geography, or complementary expertise
• Save a branch as a mission or shortlist
• No auto-generated public trail without review in the MVP

Acceptance Criteria

• Themes and profile inclusion are reproducible from public criteria
• Users can inspect the criteria
• No claims extend beyond cited evidence
• Visited state persists locally or for the authenticated user
• Mobile and keyboard navigation work
• Analytics cover trail start, branch, completion, and save
• The experiment can be disabled with a feature flag

Privacy and Trust Constraints

• Diversity comes from transparent skill, geography, and activity dimensions, not inferred sensitive attributes
• Avoid popularity-only recommendation loops

Success Metrics

• Start-to-completion rate
• Branches explored
• Profiles opened
• Missions or shortlists created
• Return rate compared with normal browsing

Related Issues

• Agent-aware developer discovery and filters #73
• Visualize the consent-based Agent Network and connection lifecycle #138