performance-review-drafter

Overview

Generates structured performance review drafts grounded in specific, verifiable evidence. Reads source materials (project docs, peer feedback, 1:1 notes, prior reviews), extracts accomplishments in STAR format, identifies growth areas with constructive framing, and produces a complete draft calibrated to the employee's role and level. Every claim in the output traces back to a cited source.

Workflow

<Definition - STAR Format> A structured method for describing accomplishments:

• Situation: The context or challenge faced
• Task: The specific responsibility or objective
• Action: What the person did, with concrete details
• Result: The measurable outcome or impact delivered

Every accomplishment in the review draft must follow this structure. Partial evidence (missing Result, for example) should be flagged for the user to complete rather than filled with assumptions. </Definition - STAR Format>

<Definition - Growth Areas> Specific skills, behaviors, or competencies where the employee has room to develop further. Growth areas are forward-looking opportunities, not failures. They describe what leveling up looks like at the next stage of the employee's career. Frame as "To reach the next level, focus on..." rather than "Failed to..." </Definition - Growth Areas>

<Definition - Performance Gaps> Concrete instances where delivered work fell short of the role-level expectations. Unlike growth areas, performance gaps reference specific shortfalls with evidence. These require careful, factual framing and must never rely on hearsay or unverified claims. </Definition - Performance Gaps>

<Definition - Evidence Quality Tiers> Rating system for source reliability:

• Tier 1 (Strong): Quantitative metrics, shipped deliverables, documented project outcomes
• Tier 2 (Solid): Peer feedback with specific examples, manager observations from 1:1 notes
• Tier 3 (Supporting): General sentiment, aggregated survey results, secondhand accounts

The draft prioritizes Tier 1 and 2 evidence. Tier 3 evidence is used only to reinforce patterns established by stronger sources. </Definition - Evidence Quality Tiers>

<Definition - Level-Calibrated Expectations> The standard against which contributions are measured depends on the role level:

• Junior (L4/equivalent): Completes assigned work reliably, asks good questions, grows technical skills
• Mid (L5/equivalent): Owns end-to-end features, mentors juniors, identifies problems proactively
• Senior (L6/equivalent): Drives cross-team initiatives, influences technical direction, multiplies team output
• Principal+ (L7+/equivalent): Shapes organizational strategy, solves ambiguous problems at scale, builds lasting systems

Always calibrate praise and growth areas to the employee's current level, not a generic standard. </Definition - Level-Calibrated Expectations>

<Workflow - Performance Review Draft description="End-to-end performance review drafting flow." tools=[file_read] triggers=["write a performance review", "draft my self-assessment", "prepare review for direct report", "perf review", "annual review draft", "write peer feedback"]

1. [Ask user] Gather any missing required inputs: employee_name, review_period, role_level, review_type. If evidence_sources were not provided, ask what materials are available (project docs, 1:1 notes, peer feedback, prior reviews, metrics dashboards). Ask about any length constraints or organizational frameworks (leadership principles, competency models) that should structure the review. If fails: Re-ask for the specific missing inputs; if the user cannot provide the required fields, stop and explain the draft cannot proceed without them.

2. [Agent] If evidence_sources were provided as file paths, read each file. Extract key information: accomplishments, metrics, feedback quotes, project outcomes, behavioral observations. Tag each extracted item with its source file and approximate date within the review period. If fails: Report which files could not be read, and ask the user to re-share the correct paths or paste the content directly.

3. [Think] Assess evidence coverage across the review period. Check for temporal distribution (early, mid, late period). Check for breadth across different competency areas (technical delivery, collaboration, leadership, communication). Identify gaps where evidence is thin or absent.

4. [Ask user] If evidence gaps exist, present them clearly. For example: "I have strong evidence for technical delivery but limited material on cross-team collaboration. Do you have additional sources, or should I note this as an area where more evidence is needed?" If recency bias is detected (per Gotchas), flag it explicitly. If fails: If the user does not respond or cannot supply more sources, note the gaps in the draft and proceed with the evidence available.

5. [Agent] Organize extracted evidence into STAR-format accomplishments. Group by theme or competency area. Rate each item using the Evidence Quality Tiers definition. Discard Tier 3 evidence that lacks Tier 1 or 2 corroboration. If fails: If evidence cannot be mapped to STAR structure, flag the incomplete items for the user to complete rather than filling them with assumptions.

6. [Think] Calibrate each accomplishment against Level-Calibrated Expectations. Determine which items represent strong performance at the employee's level, which represent baseline expectations, and which demonstrate above-level impact. Identify patterns that suggest growth areas.

7. [Think] Draft growth areas using the Growth Areas definition. Ensure each growth area is (a) supported by observable evidence or absence of evidence, (b) framed constructively, (c) paired with a specific development recommendation. Verify no growth area violates Rules 5, 6, or 9.

8. [Agent] Compose the full review draft using the Review Draft template. Adapt voice and person based on review_type (first person for self_assessment, third person for manager_review, observational third person for peer_feedback). Ensure every evaluative statement has at least one specific example. If fails: Report the composition error to the user and retry once; if it persists, present the structured evidence gathered so far so no work is lost.

9. [Agent] Run a quality check on the draft:

   • Verify no fabricated metrics or outcomes (Rule 2)
   • Confirm balanced tone (Rule 4)
   • Check temporal distribution of examples (Rule 5)
   • Confirm no named peer attribution without approval (Rule 6)
   • Confirm no comparisons to named individuals (Rule 7)
   • Verify level-appropriate calibration (Rule 8)
   • Scan for protected-class information (Rule 10)
   • Confirm no superlatives without sourced quotes (Rule 13) Flag any issues found and revise before presenting. If fails: If the quality check cannot complete, present the draft with an explicit note listing which checks were not verified.

10. [Ask user] Present the complete draft. Highlight any sections where evidence was thin and your confidence is lower. Ask the user to review for accuracy, add missing context, adjust tone, or approve. Offer to iterate on specific sections. If fails: If the draft cannot be presented, provide the draft text directly and ask the user to confirm they received it before closing.

</Workflow - Performance Review Draft>

<Template - Review Draft>

Performance Review: {{employee_name}}

{{review_period}} | {{role_level}}

Summary

[2-3 sentence overview of the employee's performance during the review period. State the overall trajectory: exceeding expectations, meeting expectations, or developing toward expectations. Ground in the single strongest evidence point.]

Key Accomplishments

[Accomplishment Theme 1]

Situation: [Context and challenge] Task: [Specific responsibility] Action: [What was done, with concrete details] Result: [Measurable outcome or impact] Source: [file/document reference]

[Accomplishment Theme 2]

[Repeat STAR format]

[Accomplishment Theme 3]

[Repeat STAR format]

Strengths

[2-4 behavioral or technical strengths demonstrated consistently across multiple examples. Each strength statement includes at least one specific instance.]

Growth Areas and Development Recommendations

[Growth Area 1]

Observation: [What the evidence shows] Opportunity: [What leveling up looks like] Recommended action: [Specific, actionable next step]

[Growth Area 2]

[Repeat format]

Overall Assessment

[1-2 paragraphs synthesizing the review. Restate key contributions, acknowledge growth trajectory, and frame development areas as investments in the employee's continued progression. End with a forward-looking statement about expectations for the next period.]

---

Draft generated from: [list of source files used] Review type: {{review_type}} This is a draft for your review and editing. Verify all claims against your direct knowledge before submitting. </Template - Review Draft>