Feedback

Workflow

1. GATHER Context

• Read manage_user_setup(action="read") to load the user's equipment, basket size, and preferences.
• Read manage_coffee(action="list") to check for existing coffee tracking files, then load the relevant one via manage_coffee(action="read", coffee_name="...").
• If no coffee context exists: ask the user what coffee they're brewing before proceeding.
• Stale check: If roast date is 30+ days old, gently ask if user is still on this bag.

2. COLLECT Feedback

Gather from the user (ask for what's missing):

Minimum viable feedback: Rating + balance + one specific observation.

Dose in & dose out: The brew ratio (dose in : dose out) is important for diagnosis and tracking. Make sure both are captured:

• For dose out: if a BT scale is connected, check final_weight_g in analyze_shot output first. If telemetry is unavailable or looks unreliable, ask the user.
• For dose in: ask if not volunteered — it should match the basket size from manage_user_setup.
• If the user doesn't mention either, ask casually: "What were your dose in and output weight?"
• Log both in shot notes via dose_in and dose_out parameters.

Scale data handling: Prefer telemetry for dose out when a shot ID is available (see read_knowledge(action="read", filename="diagnostics/TELEMETRY_PATTERNS") for scale artifact detection). If telemetry is unavailable or looks unreliable, just ask the user.

3. ANALYZE & RECOMMEND

Load these knowledge files for diagnosis:

• read_knowledge(action="read", filename="ESPRESSO_BREWING_BASICS") — variable hierarchy (grind → ratio → temp → pressure → puck prep) and diagnostic decision tree
• read_knowledge(action="read", filename="ESPRESSO_TASTING_GUIDE") — taste diagnosis and the Scott Rao channeling rule (sour + bitter = channeling → fix puck prep, NOT grind)
• read_knowledge(action="read", filename="diagnostics/DIAGNOSTIC_TREES") — full diagnostic decision trees for complex cases

Apply the diagnostic rules to the user's feedback. Always explain why you're suggesting a change. One primary recommendation, one backup.

4. RECORD

Do all of these automatically after feedback is collected:

4a. Shot Notes → Device (via MCP)

If a shot ID is available, sync feedback to the device:

manage_shot_notes(shot_id, action="update", rating=X, balance_taste="...", notes="...", grind_setting="...", dose_in=X, dose_out=X)

4b. Coffee Tracking (via MCP)

Log a brewing journal entry to the coffee's tracking file — analysis, not raw numbers:

manage_coffee(
  action="log_entry",
  coffee_name="[coffee-name]",
  entry_date="YYYY-MM-DD",
  entry_headline="Grind [X], [Profile Name] — [X/5]",
  entry_body="[Agent's analysis: what worked, what didn't, taste description, what to try next. Reference shot ID if available.]"
)

The journal entry should capture thinking, not numbers. Include:

• What the shot tasted like and why (connect to extraction theory)
• What worked well and what didn't
• Specific recommendation for next shot
• Shot ID reference if available (for analyze_shot later)

If no coffee tracking file exists yet, create one first:

manage_coffee(
  action="create",
  coffee_name="[coffee-name]",
  roaster="...",
  origin="...",
  process="...",
  roast_level="...",
  roaster_notes="[tasting notes from bag]",
  approach="[Profile name] at [temp]. [Pressure logic reasoning]. Starting at grind [X], [dose]g in, targeting 1:[ratio]. [Why this approach suits this bean.]"
)

4c. Grind Map (if 4+ stars)

If 4+ stars AND grind setting provided, add to the grind map:

manage_grind_map(
  action="add_entry",
  coffee="[name]",
  roast="[light/medium/dark]",
  process="[washed/natural/honey/anaerobic]",
  origin="[country/region]",
  days_off_roast="[X days]",
  grind="[setting]",
  profile="[name]",
  ratio="1:X",
  temp="X°C",
  rating="X/5",
  date="YYYY-MM-DD"
)

4d. Brewing Insights (when patterns emerge)

When a meaningful pattern emerges (not after every shot), update the cross-coffee insights:

1. Read current insights: manage_brewing_insights(action="read") (or init via manage_brewing_insights(action="init") if it doesn't exist)
2. Update with the new learning:

manage_brewing_insights(
  action="write",
  content="[updated full markdown with the new insight added to the appropriate section]"
)

When to update brewing insights:

• Coffee is dialed in (4+ stars, balanced) — record what worked and why
• A clear pattern emerges (e.g. "declining pressure works for all medium-roast honeys we've tried")
• A surprising finding worth remembering across coffees

When NOT to update: After every single shot. Only when there's a genuine cross-coffee learning.

5. SUGGEST Next Steps

Based on the analysis:

If still dialing in (rating < 4 or not balanced):

• State the specific change for the next shot
• Explain what to watch for ("Time to first drip should increase" / "Look for more body")

If dialed in (rating 4+ AND balanced):

• Celebrate briefly
• Recommend a drink format based on shot character:

Core principle: Extract for the bean's best expression first, then match the drink format. Never adjust grind/ratio/pressure/temp to "make the shot work in milk."

If user wants full milk science, steaming technique, or drink recipes → reference MILK_AND_DRINKS.md knowledge file.

Integration with Other Skills

• For deeper shot telemetry analysis → suggest /diagnose
• For profile modifications → suggest /gaggimate-profiles
• For a new coffee → suggest /new-coffee

Quick Reference

User says: "3 stars, sour, grind 12, 22g in" Action: Gather context → analyze → record journal entry (analysis, not numbers) → diagnose (sour = extract more) → recommend grind/yield change

User says: "5 stars, balanced, amazing sweetness" Action: Gather context → celebrate → record journal entry → add to grind map → update brewing insights if pattern emerged → recommend drink format

User says: "it was sour AND bitter" Action: Gather context → diagnose channeling → recommend puck prep fix, NOT grind change → record journal entry