A Claude Code-powered tool for crafting thoughtful, comprehensive performance reviews (CONNECT reviews) at Octopus Deploy.

This tool helps managers:

1. Gather and organize all relevant information about an employee's performance
2. Work through conversations (structured or ad-hoc) to clarify their thoughts
3. Generate complete CONNECT review write-ups with proper framing and goal-setting
4. Reference role definitions and competencies from Octopus People

.
├── README.md                    # This file
├── inputs/                      # Employee-specific materials (GIT IGNORED)
│   └── employee-name/          # Example template - copy for each employee
│       └── braindump.md        # Quick-start guide and notes file
├── framework/                   # What shapes reviews (committed to git)
│   ├── connect-template.md     # CONNECT review questions
│   ├── output-guidelines.md    # How outputs should be formatted
│   └── octopus-people/         # Role definitions (git submodule)
└── outputs/                     # Generated reviews (GIT IGNORED)
    └── [employee-name].md      # Single file per employee

1. Install git hooks (protects employee data from being pushed to GitHub):

   ./setup-hooks.sh

   This prevents accidentally pushing review branches to GitHub. Only main branch can be pushed.

2. Update Octopus People data (optional, already cloned):

   cd framework/octopus-people
   git pull origin main
   cd ../..

   See UPDATE_OCTOPUS_PEOPLE.md for details.

3. Start Claude Code in this directory:

   claude

1. Create input directory by copying the example:

   cp -r inputs/employee-name inputs/jane-smith

2. Add materials to inputs/jane-smith/:

   • Edit braindump.md to add your thoughts
   • Add other files: 1:1 notes, project notes, peer feedback, etc.
   • File naming is flexible - whatever works for you

3. Work on the review using Claude Code:

   Option A: Structured Interview

   /connect-interview jane-smith
   

   Option B: Incremental Conversation (more common)

   Let me work on Jane's review
   Jane did excellent work on the API redesign project...
   

   Option C: Direct commands

   Review what I have for Jane so far
   Let's work on the technical leadership section
   

4. Find your output at outputs/jane-smith.md

Managers work on reviews over days or weeks, not in one sitting. This tool supports:

• Drop in anytime: Add a thought, Claude updates the review
• Structured interviews: Use /connect-interview when you want guided questions
• Gap analysis: "Review what I have and tell me what's missing"
• Focused work: "Let's work on the mentoring section"
• Direct refinement: "Add more examples to the technical leadership part"

Claude maintains the output file across sessions, building it up incrementally.

Reviews are generated as single markdown files (outputs/[employee-name].md) containing:

• Complete answers to all CONNECT template questions
• Direct quotes from role definitions for every strength/growth area
• Inline goals within the growth areas
• Actionable "how" guidance for development areas

See framework/output-guidelines.md for detailed formatting requirements.

This repo uses a two-branch strategy to protect employee data:

main branch (pushed to GitHub):

• Tool structure: README, framework, slash commands, documentation
• framework/octopus-people/ - Public organizational data
• Example template at inputs/employee-name/
• ❌ NO employee review data

sensitive-data branch (local only, NEVER pushed):

• Everything from main, PLUS your actual review data
• inputs/[employee-name]/ - Your notes and materials
• outputs/[employee-name].md - Generated reviews
• ✅ Versioned locally via git for safety
• 🛑 Blocked from pushing to GitHub by pre-push hook

1. Install protection: Run ./setup-hooks.sh after cloning (one-time setup)
2. Work on reviews: Claude automatically uses sensitive-data branch
3. Tool updates: Pulled from main branch on GitHub
4. Your data: Safe on local sensitive-data branch, never pushed

The pre-push hook ensures you cannot accidentally push review data to GitHub.

• framework/output-guidelines.md - Critical reference for what outputs should look like
• framework/connect-template.md - CONNECT review questions to answer
• inputs/employee-name/braindump.md - Example/template for getting started
• claude.md - Instructions for Claude Code (always in context)

This is a living tool. Feel free to update documentation, templates, and workflows as needed!