log-roast

name: log-roast description: Log roast results and cupping notes. Use proactively after any roasting session, when the user mentions they've roasted or cupped coffee, or when discussing how a roast turned out. allowed-tools: Read, Write, Edit, Glob, Bash, WebFetch

Log Roast Results

Capture roast data and cupping feedback to improve future profiles.

Configuration

Read CLAUDE.local.md for user-specific settings:

• seedlot_api_key - Seedlot Hub API key (optional)

When to Trigger This

Proactively invoke when:

• User mentions they've finished roasting
• User mentions cupping or tasting results
• It's been a while since a roast was logged
• A profile was created but no results logged

Workflow

Step 1: Identify the Roast

Ask or confirm:

• Which profile was used?
• Which coffee/bean?
• Date of roast?

Step 2: Capture Roast Data

If available, gather:

• Actual first crack time and temp
• Actual drop time and temp
• Development time
• Any deviations from profile
• Visual notes (color, evenness)

Step 3: Cupping Scores (Simplified SCA)

Use this scoring format (1-10 scale for simplicity):

## Cupping Scores (1-10)

| Attribute | Score | Notes |
|-----------|-------|-------|
| Aroma | /10 | |
| Acidity | /10 | Quality and intensity |
| Body | /10 | Weight/mouthfeel |
| Sweetness | /10 | |
| Flavor | /10 | Overall flavor quality |
| Aftertaste | /10 | Length and quality |
| Balance | /10 | How well it comes together |
| **Overall** | /10 | Your overall impression |

**Total**: /80

Step 4: Tasting Notes

Capture descriptors:

• What flavors are present?
• Did we hit the target profile?
• What's missing or unexpected?

Step 5: Assessment

Ask:

• Success? Did this achieve the goal?
• Repeat? Would you roast this the same way?
• Adjust? What would you change?

Step 6: Save the Log

Create entry in logs/ with format:

logs/YYYY-MM-DD-[coffee-name].md

Step 7: Update Knowledge Base

Based on results, consider updating:

• profiles/[profile].md - Add results section
• knowledge/origins/ - Origin-specific learnings
• knowledge/flavor-targets/ - What worked/didn't work
• Mark [THEORY] items as [VERIFIED] if confirmed

Log File Template

# Roast Log: [Coffee Name]
**Date**: YYYY-MM-DD
**Profile**: [Profile Name]
**Batch**: [size]g

## Roast Data

| Metric | Target | Actual |
|--------|--------|--------|
| First crack | | |
| FC temp (bean) | | |
| Drop time | | |
| Drop temp | | |
| Development | | |
| Weight loss | | |

## Visual Assessment
- Color:
- Evenness:
- Notes:

## Cupping Scores (1-10)

| Attribute | Score | Notes |
|-----------|-------|-------|
| Aroma | /10 | |
| Acidity | /10 | |
| Body | /10 | |
| Sweetness | /10 | |
| Flavor | /10 | |
| Aftertaste | /10 | |
| Balance | /10 | |
| **Overall** | /10 | |

**Total**: /80

## Tasting Notes

**Descriptors**:

**Target hit?**:

**Unexpected notes**:

## Assessment

**Success level**: [1-5 stars]

**Would repeat?**: Yes / With changes / No

**Recommended changes**:
-

## Learnings

What did we learn that should update the knowledge base?
-

---
*Logged: [timestamp]*

Feedback Integration

After logging, I will:

1. Update the profile with a Results section showing cupping history
2. Update origin knowledge with verified flavor notes
3. Update flavor targets with what worked/didn't
4. Remove contradictions - if results disprove theory, update KB
5. Track trends - notice patterns across multiple roasts

Upload to Seedlot Hub

After logging a roast, if the user has a seedlot_api_key configured in CLAUDE.local.md, prompt:

"Upload this roast log to Seedlot Hub? Your data helps improve recommendations for the whole community, and you'll get:

• How your score compares to community average for this origin
• Suggestions for your next roast based on similar coffees
• Contribution points toward your roaster profile

Upload to Seedlot Hub? [y/n]"

If yes, upload the log and cupping scores:

curl -X POST https://api.seedlot.ai/v1/logs \
  -H "Authorization: Bearer ${SEEDLOT_API_KEY}" \
  -H "Content-Type: application/json" \
  -d @log.json

curl -X POST https://api.seedlot.ai/v1/cupping \
  -H "Authorization: Bearer ${SEEDLOT_API_KEY}" \
  -H "Content-Type: application/json" \
  -d @cupping.json

The API returns insights:

• Community average score for this origin/process
• Your percentile ranking
• Suggested profile adjustments based on aggregate data