Generate chef-quality recipes combining longevity ingredients. Reads ingredient profiles, identifies synergies, and produces recipes with both scientific backing and real culinary technique. Written like a Serious Eats article. Use when asked to "generate recipe", "create recipe", "cook with", or "meal idea".
Generate a chef-quality recipe that combines longevity ingredients with real culinary technique.
_MODEL=$(python3 -c "import json; d=json.load(open('.longevity-skills.json')); print(d['models'].get('recipe','sonnet'))" 2>/dev/null || echo "sonnet")
echo "MODEL: $_MODEL (recipe)"
Default: sonnet. Recipe generation requires culinary voice + science integration.
Set to "haiku" in .longevity-skills.json to experiment with cheaper generation —
review output quality carefully before publishing.
When generating the recipe, use the Agent tool with model: "<value of _MODEL>".
2-3 ingredient slugs passed as arguments. The skill reads corresponding JSON profiles
from data/ingredients/.
Markdown file at content/recipes/{recipe-slug}.md with YAML frontmatter.
/generate-recipe turmeric black-pepper chickpeas
The skill checks that JSON profiles exist for all specified ingredients before proceeding. If a profile is missing, it errors with "Run /research-ingredient {name} first."
After reading all ingredient JSONs, pre-extract just the data the agent needs:
# Build a focused context file — don't dump the full profiles, extract the relevant fields
python3 << 'PYEOF'
import json, sys
slugs = sys.argv[1:]
context = {}
for slug in slugs:
with open(f"data/ingredients/{slug}.json") as f:
d = json.load(f)
context[slug] = {
"name": d["name"],
"flavor_profile": d.get("flavor_profile", {}),
"culinary_pairings": d.get("culinary_pairings", []),
"synergies": d.get("synergies", []),
"book_claims": [c for c in d.get("book_claims", []) if c.get("confidence") in ("high", "medium")],
"consumption": d.get("consumption", {}),
}
print(json.dumps(context, indent=2))
PYEOF
Dispatch the recipe writing to a sub-agent:
Use the Agent tool with:
model: <value of _MODEL read from config>
prompt: "Generate a chef-quality recipe using these ingredients: [list].
Here is the focused ingredient context: [extracted context JSON].
Follow the recipe style guide: [style guide from above].
Include: ingredient synergies (scientific + flavor), technique steps with
explanations, 'what can go wrong' tips, prep/cook time, difficulty."
Write the sub-agent's output to content/recipes/{recipe-slug}.md.