EqualsGroup/assess

Instructions

Run the performance assessment comparison workflow for a named report.

Phase 1: Check existing files

Check $EA_ROOT/outputs/assessments/[name]/ for:

• manager-assessment-YYYY.md
• self-assessment-YYYY.md

Use current year. Directory uses the same [first-last] format as people files (e.g., outputs/assessments/jane-doe/).

Phase 2+3: Fetch missing + read context (parallel)

Fetch missing assessments (if needed):

• Self-assessment missing: Use ~~hr-system connector to navigate to the employee's review page, extract content, save to outputs/assessments/[name]/self-assessment-YYYY.md
• Manager assessment missing: Ask the user to provide it

Read memory context (parallel):

• Both assessment files
• $EA_ROOT/memory/people/[name].md
• $EA_ROOT/memory/dynamics.md
• $EA_ROOT/memory/meetings.md — their 1:1 prep notes

Phase 4: Analysis

Produce a structured comparison:

## Assessment Comparison: [Name] — [Year]

### Self-Assessment Highlights
- [Key points from their self-assessment — what they're proud of, where they think they struggled]

### Alignment (both agree)
- [Areas where manager and self-assessment see the same strengths/weaknesses]

### Key Divergences
- **[Topic]** — Self: "[quote/summary]" vs Manager: "[quote/summary]"
  - *Implication:* [What this means for the review conversation]

### Perception Gaps
- **Feeling valued:** [Their score and exact words]
- **Engagement:** [Self-perception vs manager rating]
- **Biggest blind spot:** [What they don't see about themselves]

### Coaching Points for Review Conversation
- [What to reinforce — strengths to explicitly acknowledge]
- [What to develop — frame with empathy, connect to their goals]
- [What to monitor — areas to track without over-indexing]

### Conversation Approach
- **Tone:** [e.g., "supportive but direct", "softer — reframe around growth"]
- **Open with:** [Suggested opening]
- **Avoid:** [Landmines to steer around]
- **Key message to land:** [The one thing they should walk away understanding]

Phase 5: Update memory files

1. Person's file — add/update a **YYYY Performance Review Insights:** section with:
   • Self-assessment highlights
   • Alignment with manager review
   • Key divergences (with specific quotes/evidence)
   • Coaching points for the review conversation
   • Review conversation approach notes
2. meetings.md — update 1:1 prep notes with review context
3. my-work.md — add or update "conduct performance review conversation" item

Phase 6: Save output

Save to $EA_ROOT/outputs/assessments/[name]/comparison-YYYY.md

Trigger Rules

When a self-assessment completion notification arrives (e.g. from the HR system):

1. Check if $EA_ROOT/outputs/assessments/[name]/self-assessment-YYYY.md already exists
2. If not, fetch the self-assessment content and save it there
3. If both manager and self assessment now exist, run the Assessment Comparison Workflow (Phase 4+5+6 above)

Checking assessment status: Before asking the user whether reviews are complete, always check $EA_ROOT/outputs/assessments/ to see what files exist. The presence of files tells you the answer.

Fetching assessments from HR system: Use Claude's own Chrome integration (mcp__claude-in-chrome__* tools) to navigate to the HR system (see $EA_ROOT/memory/company.md > Service URLs for the URL) and fetch assessment content. Do NOT use other Chrome MCPs (e.g. mcp__chrome-devtools__*) as they don't share browser sessions/cookies. This is an ad-hoc browser task, not part of the /sync flow.

Rules

• Never fabricate assessment content
• Quote directly from assessments for divergences
• Be psychologically astute — identify what the person needs to hear vs wants to hear
• Consider dynamics.md for salary frustrations, flight risk signals
• If both assessments exist and comparison was done, ask what to revisit