Performance Review Assistant

Scenario: Performance reviews in your organisation are inconsistent. Some managers write detailed, actionable feedback. Others write vague platitudes. The result: reviews don't drive development, don't differentiate performance fairly, and don't hold up under scrutiny. You want every manager using the same structured, rubric-driven approach.

Patterns used:

• RubricDesigner — designs objective, behavioural criteria for performance dimensions
• FeedbackComposer — structures constructive, specific, developmentally useful feedback
• FormativeAssessmentFramework — frames the review as a development input, not just a rating

Integration: Blueprint — one template, consistent output across the whole organisation

---

import mycontext

from mycontext.structure import Blueprint
from mycontext.foundation import Guidance, Constraints
from mycontext.templates.enterprise.evaluation import RubricDesigner, FormativeAssessmentFramework
from mycontext.templates.enterprise.communication import FeedbackComposer
from mycontext.intelligence import QualityMetrics, TemplateIntegratorAgent
from pathlib import Path

metrics = QualityMetrics(mode="heuristic")

review_blueprint = Blueprint(
    name="performance_review",
    guidance=Guidance(
        role="People operations expert and executive coach writing performance reviews",
        rules=[
            "Every strength and development area must include a specific behavioural example",
            "Feedback must be developmentally useful: 'do more of X' not 'be better at X'",
            "Rating must follow directly from evidence — never from gut feel",
            "Development goals must be SMART: specific, measurable, achievable, relevant, time-bound",
            "Tone must be fair and honest — neither artificially positive nor punitive",
        ],
    ),
    directive_template=(
        "Write a performance review for:\n\n"
        "Name: {employee_name}, Role: {role}\n"
        "Review period: {review_period}\n"
        "Manager notes: {manager_notes}\n"
        "Self-assessment: {self_assessment}\n"
        "Key outcomes delivered: {outcomes}\n"
        "Behavioural examples: {examples}"
    ),
    constraints=Constraints(
        must_include=[
            "rating with justification",
            "3+ specific strengths with examples",
            "2+ development areas with specific suggestions",
            "SMART development goals",
            "career conversation starter",
        ],
        format_rules=["## headings for each section", "Rating at the top"],
    ),
    token_budget=3000,
    optimization="quality",
)


def generate_performance_review(review_data: dict, save_path: str | None = None) -> str:
    # Integrate all three patterns first for richer analysis
    integrator = TemplateIntegratorAgent(provider="openai")
    pre_analysis = integrator.integrate(
        templates=[RubricDesigner, FormativeAssessmentFramework, FeedbackComposer],
        input_data={
            "concept": f"Performance review for {review_data['role']}",
            "learner_level": "professional",
            "context_section": f"Manager notes: {review_data['manager_notes'][:500]}",
        },
    )

    ctx = review_blueprint.build(**review_data)
    ctx.knowledge = pre_analysis.context.execute(
        provider="openai", model="gpt-4o-mini"
    ).response[:1000]

    score = metrics.evaluate(ctx)
    print(f"Context quality: {score.overall:.0%}")

    review = ctx.execute(provider="openai", model="gpt-4o").response

    if save_path:
        Path(save_path).write_text(
            f"# Performance Review\n**{review_data['employee_name']} | {review_data['role']}**\n\n{review}"
        )

    return review


review = generate_performance_review(
    review_data={
        "employee_name": "Jordan Lee",
        "role": "Senior Software Engineer",
        "review_period": "H2 2025",
        "manager_notes": (
            "Jordan delivered the payments migration on time, a technically complex project. "
            "Very strong technically. Sometimes doesn't speak up in cross-team meetings even when they have important context. "
            "Mentored two junior engineers effectively. Can be slow to flag blockers."
        ),
        "self_assessment": (
            "I completed the payments migration and am proud of the technical quality. "
            "I want to improve my communication in large group settings and be better at proactive status updates."
        ),
        "outcomes": (
            "Delivered payments migration (Q4). Reduced processing latency 40%. "
            "Mentored 2 juniors. Contributed to 3 on-call rotations."
        ),
        "examples": (
            "Payments migration: proposed and implemented a zero-downtime approach. "
            "Mentoring: regularly pair-programmes with Ana and Ben. "
            "Cross-team meeting (Nov): had important info about API deprecation but didn't share until asked."
        ),
    },
    save_path="reviews/2025-H2/jordan-lee.md",
)
print(review)

What You Get

A complete, fair performance review with:

• Rating based directly on evidence, not impression
• Strengths: specific behaviours with concrete examples ("During the payments migration, Jordan...")
• Development areas: actionable suggestions, not vague labels ("In future cross-team reviews, proactively share X by Y...")
• SMART goals: with measurable targets and time bounds
• Career conversation opener: a question or prompt to start the development discussion

The Blueprint ensures every manager — regardless of writing skill or natural inclination — produces reviews that meet the same quality bar.