Start by clarifying the goal: to reduce noise and surface only actionable, relevant articles for the athlete and coaching team. Then structure the prompt with clear role, inclusion/exclusion criteria, output format, and safeguards against speculation. Emphasize the importance of precision over recall to avoid missing critical information.
Pro tip: Frame the prompt as a decision-support tool, not a replacement for human judgment. Mention that you would include a confidence score and a feedback loop to refine the filter over time.
Specify that the model acts as a media analyst for an Olympic champion, tasked with filtering news to support performance and well-being. Emphasize that the goal is to identify actionable insights, not to summarize all news.
List what to include (e.g., training tips, competition updates, health and recovery, mental performance, logistics) and exclude (e.g., gossip, unverified rumors, negative speculation, distractions).
Require a JSON output with fields: decision (e.g., 'include' or 'exclude'), relevance_score (0-1), category (e.g., 'training', 'health'), summary (concise), and reason (brief justification).
Instruct the model to avoid speculation, rely only on facts from the article, and flag any sensational language. Require it to state 'insufficient information' if the article lacks credible details.
Include the article's headline, source, date, and body text in the prompt, and ask the model to apply the criteria and return the structured output.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.