add CLAUDE.md — agent instructions for location file processing

This commit is contained in:
2026-03-25 17:25:29 -05:00
parent c2913d6df0
commit 79e76885fe

88
23-Cooking/CLAUDE.md Normal file
View File

@@ -0,0 +1,88 @@
# Cooking Framework — Claude Code Instructions
This directory contains a personal cooking framework for a frequent traveler. The system uses a formula-based approach: every meal = Protein + Produce + Starch + Fat + Acid + Aromatic + Technique.
## Key Reference Files
- `Flavor Sense.md` — the four flavor mechanisms (fat as solvent, acid resets palate, salt in layers, browning creates umami)
- `Flavor Profile Formulas.md` — 10 flavor profiles as role principles with "identify by" cues
- `Role Substitution Logic.md` — substitution hierarchies for each role
- `Frameworks.md` — three cooking techniques: One-Pan Sear, Steam-Sauté, Starch-Buffer
- `Weekly Planning Template.md` — market-first meal planning entry point
- `Essential Non-Perishables.md` — pantry stocking guide
---
## Location File Workflow
Location files live in `Locations/[City, Country].md`. They have four zones:
- **Zone 1 (Header):** Location name, region, first visited, last updated
- **Zone 2 (Observations):** User-filled market notes — freeform or structured fields
- **Zone 3 (Analysis):** Agent-generated output — five subsections
- **Zone 4 (Visit Log):** Chronological user notes from each visit
### When the user says "Process my [City] location file"
1. Find the file at `Locations/[City, Country].md`
2. Read Zone 2 (Observations) — parse both freeform text and structured fields; treat them as the same pool of information
3. Map observations to framework roles using `Role Substitution Logic.md`:
- Fat: any cooking oil, animal fat, or dairy fat mentioned
- Acid: citrus, vinegar, fermented liquids, sour dairy, or sour-tasting condiments
- Aromatic: alliums + any secondary aromatics (ginger, lemongrass, spices, herbs)
- Starch: any grain, root vegetable, legume, or bread
- Protein: any meat, fish, egg, legume, or tofu
- Umami: any fermented condiment, aged cheese, concentrated tomato, or mushroom
4. Match observations to the 10 profiles in `Flavor Profile Formulas.md`. Assign confidence:
- ✦✦✦ = all three defining elements (fat + acid + aromatic character) present
- ✦✦ = two of three defining elements present, or key secondary aromatics missing
- ✦ = only partial match; note specifically what's absent
- Omit profiles that score below ✦ (nothing available)
5. Generate 35 meal options using available ingredients. Each formatted as:
`*[Meal name]* — [Technique] · [Profile] · [protein] / [produce] / [starch] / [fat] / [acid] / [aromatic]`
Use techniques from `Frameworks.md`: One-Pan Sear or Steam-Sauté (Starch-Buffer if microwave available)
6. Write 35 bullet points of regional intelligence — what's culinarily notable about this location, what differs from the framework's defaults, what a traveler without local knowledge would benefit from knowing. Draw on your knowledge of the region's culinary traditions.
7. Write 35 "things to look out for" — specific ingredients or products worth seeking on the next market visit, each with a one-line explanation of why it matters (flavor role, rarity, or value)
8. Replace Zone 3 (Analysis) with the five subsections filled in
9. In Zone 1, set "First visited" to today's date if currently blank; set "Last updated" to today's date
10. Stage and commit:
```
git add "23-Cooking/Locations/[City, Country].md"
git commit -m "generate location file: [City, Country]"
```
### When the user says "Update my [City] location file"
1. Find the file at `Locations/[City, Country].md`
2. Read Zone 2 (Observations) AND Zone 4 (Visit Log) — the full visit log supplements and may revise the original observations; treat both as one combined pool of information
3. Re-run steps 38 from the generate workflow above using the combined observations
4. In Zone 1, set "Last updated" to today's date; leave "First visited" unchanged
5. Stage and commit:
```
git add "23-Cooking/Locations/[City, Country].md"
git commit -m "update location file: [City, Country]"
```
---
## Analysis Output Format
### Ingredient Role Map
Inline format, one line:
> **Fat:** [items] | **Acid:** [items] | **Aromatic:** [items] | **Starch:** [items] | **Protein:** [items] | **Umami:** [items]
### Profile Matches
One line per profile, ordered by confidence (highest first):
> [Profile Name] ✦✦✦
> [Profile Name] ✦✦ (missing: [specific absent element])
> [Profile Name] ✦ (missing: [specific absent elements])
### Meal Options
35 items, one per line, italicised meal name:
> *[Meal name]* — [Technique] · [Profile] · [protein] / [produce] / [starch] / [fat] / [acid] / [aromatic]
### Regional Intelligence
35 bullet points. Focus on: staple ingredients, local cooking patterns, things that differ from framework defaults, and anything a traveler without local knowledge would miss.
### Things to Look Out For
35 items, each with name and one-line explanation:
> **[Ingredient/product]** — [why it matters: flavor role, what profile it unlocks, rarity, or value]