What the work actually is

Each environment is a reconstruction of a comms team's tool surface at a specific moment: an embargo about to break, a reporter with a conflicting source, a corrections request sitting in an inbox next to a launch checklist nobody updated. Your job is to build the moment and the trap inside it. A brief might read "the announcement goes out Thursday" while the shared drive holds a spokesperson quote that was never legal-approved, the tracker shows an exclusive already promised to a second outlet, and the analytics dashboard shows the embargo already leaked. You specify every document, dashboard, persona and tool state the scenario requires, write the reference answer, and articulate the criteria that separate a strong response from one that is fluent and wrong. Then you review AI agent attempts and score them against your own standard.

The capability areas are stated plainly: diagnosis from messy or conflicting data, prioritization and tradeoffs, planning and execution, QA and reconciliation, triage and escalation, research and evaluation, reporting, and multi-tool orchestration. Within PR and comms that maps to media pitching, embargo and exclusive strategy, announcement sequencing, interview prep, coverage evaluation, corrections, and launch communications.

What the screen is looking for

  • Hands-on, not adjacent. The listing is explicit: you have pitched reporters, run an embargo, and handled a correction yourself. Agency oversight or approving someone else's pitch list reads differently under follow-up.
  • Reasoning you can write down. Much of the value is in the explanation of why a decision is right — a scored rubric that only says "good judgment" is unusable.
  • Environment-building instinct. Prior work on AI training environments, RL environments or simulated case studies is strongly preferred, as is Rubric Academy or Rubric Bootcamp certification. Neither is stated as mandatory.

Applicants complete a short multiple-choice knowledge screener specific to PR and communications before human review.

Logistics

Remote and asynchronous. Observed pay for this band runs $60–100/hr, set by depth of experience and screener performance — as with all contract evaluation work, that is what has been reported, not a guarantee. Hours are flexible and self-directed; contributors typically commit somewhere between a few hours a week and something closer to part-time, and work is usually structured as batches of environment design or agent reviews with a delivery deadline rather than scheduled shifts.