Performance Review Drafter

What it takes in

The inputs you provide on each run, validated before anything executes.

2 tasks, run in order.

Each task passes its result to the next. Here's what each does and what it produces.

1. 01
   Draft the review. From {{achievements}} and, if provided, {{managerNotes}}, write a complete performance review draft: an overall summary of the cycle, strengths each tied to specific evidence from the inputs, growth areas framed constructively (behavior, impact, and the change to make), a goal-by-goal assessment of {{previousGoals}} (met, partially met, or not met, with the supporting evidence; if no goals were provided, note that and assess against the achievements and observations instead), and 3 to 5 suggested goals for the next cycle that follow from the growth areas and the trajectory shown. Do not fabricate achievements, metrics, or incidents not present in the inputs. Present readably, never JSON.

   Expected output

   A readable review draft with no JSON and no code blocks containing: Summary (a short paragraph on the overall cycle), Strengths (bullets, each pairing the strength with its specific evidence), Growth Areas (bullets, each framed as behavior, impact, and suggested change), Goal Assessment (one entry per previous-cycle goal marked Met, Partially Met, or Not Met with evidence, or a note that no goals were provided), and Suggested Next-Cycle Goals (3 to 5 specific, measurable goals).
2. 02
   Run the fairness pass. Review the draft from Task 1 for three failure modes: vague claims that lack supporting evidence from the inputs, recency bias (feedback that over-weights recent events while ignoring earlier parts of the cycle covered in the notes), and loaded or biased language (character judgments, gendered descriptors, or phrasing like 'not a culture fit'). List every issue found with the original wording and the fix applied, then show the complete corrected final version of the review with all fixes incorporated. If nothing needed fixing, say so and present the final version unchanged. Present readably, never JSON.

   Expected output

   A readable fairness report with no JSON and no code blocks containing: Fairness Flags (bullets grouped under Vague Claims, Recency Bias, and Loaded Language, each showing the original wording and the corrected wording, or a note that the draft was clean) and Final Review (the complete corrected review with the same sections as Task 1, ready to paste into a review tool).