Setup Instructions

# Company Context — Setup Instructions

This directory is the customization layer for Performance Review.
Fill in the files below to tailor the skill to your organization.
The skill automatically detects filled vs. placeholder content and adjusts accordingly.

## How It Works

- **Placeholders only** (`[Fill in: ...]`): The skill uses universal engineering leadership frameworks.
- **Partially filled**: The skill applies filled fields and ignores remaining placeholders. For example, you can define company values without filling in the full IC ladder.
- **Fully filled**: Company context takes precedence over generic defaults in all output.

## Files in This Directory

| File | What to Fill In |
|---|---|
| `ladder.md` | Company values with behavioral anchors, IC level expectations, manager competencies, and promotion criteria |
| `process.md` | Review cadence, timelines, rating scales, calibration format, tooling, and cultural norms |
| `templates.md` | Required output formats for standard and milestone reviews |

## What Changes When You Fill This In

When company context is present, the skill evaluates employees against your specific values and competencies (not just generic engineering expectations), uses your rating scale and format, follows your review timeline, and adapts output to match your tool's required fields. The more files you fill in, the more tailored the reviews become.

## Routing Logic (for reference)

Claude applies company context using this priority:
1. **Direct Match**: If the user specifies a review type that matches a milestone defined in `process.md`, use the milestone-specific scale and template as the primary framework. Use generic `references/` files as supportive knowledge.
2. **Hybrid Match**: For standard reviews, use the generic `references/` frameworks as the primary structure, but pull company values from `ladder.md` as a mandatory cultural rubric applied to every evaluation.
3. **Placeholder Fallback**: If a company section still contains `[Fill in: ...]` placeholders, ignore it and use the detailed expertise in the `references/` knowledge files.

## Required Maintenance

- **Metric consistency**: Any new competencies or key results added to these files should use observable, behavioral language — not personality traits or abstract adjectives.
- **Formatting**: Use clean Markdown tables and headers. Do not use XML tags — they will break the skill's parsing.