Evan
102b506b0b
All 32 agent manifests and 17 hands shipped with empty mcp_servers /
skills lists, which the kernel interprets as "no filter" — every
globally-configured MCP server's tools and every installed skill get
injected into the prompt on every LLM call. On a typical instance (9
MCP servers, ~85 MCP tools + ~82 built-in tools) that's ~50k input
tokens per turn spent on definitions the agent never uses.
Changes
-------
32 agents/*/agent.toml:
- mcp_servers: 1-4 per agent. memory wherever state persists across
turns; fetch / exa-search / brave-search only where the prompt
actually calls for web; git / github / filesystem on engineering
agents; gmail / google-calendar / linear / jira on productivity
agents whose prompts mention them.
- skills: per-role allowlist driven by what the system_prompt names
(e.g. coder → rust/python/typescript/git/shell-scripting; devops-
lead → docker/kubernetes/terraform/ansible/ci-cd/helm/prometheus/
sysadmin). Generalists (assistant) keep skills = [] (see "Open
items" below).
- skills_disabled = true on the four short-conversational agents
(hello-world, recipe-assistant, health-tracker, home-automation).
Their system prompts never instruct the LLM to consult any skill,
so loading all 60 was pure waste. They also drop the explicit
max_history_messages override and inherit the kernel default (60).
- max_history_messages tiered by workload shape:
60 short conversational (hello-world, recipe, health-tracker,
home-automation) — inherits the rising kernel default
(`DEFAULT_MAX_HISTORY_MESSAGES = 60`); no override needed.
60 single-turn task agents (writer, translator, doc-writer,
email-assistant, customer-support, sales-assistant, recruit-
er, social-media, personal-finance, tutor, travel-planner,
meeting-assistant, ops, devops-lead, planner) — explicit
override at the same value to lock the cap if the kernel
default moves again.
80 multi-step / tool-heavy (coder, debugger, architect, code-
reviewer, test-engineer, security-auditor, analyst, data-
scientist, academic-researcher, researcher, legal-assistant)
120 coordinators (assistant, orchestrator) — long multi-agent
sessions where prompt-cache continuity is critical
All values sit at or above the kernel default. Pinning lower
would thrash the prompt cache (the failure mode #91 fixed for
the creator hand by *raising* the cap, not lowering it).
17 hands/*/HAND.toml:
- hand-level mcp_servers / skills now declared on every hand, so
every [agents.*] inside inherits a sensible allowlist.
- skills_disabled = true placed on each [agents.*] inside clip and
creator (pure media pipelines that don't benefit from any skill).
HandDefinitionRaw in librefang-hands does NOT have a top-level
skills_disabled field — declaring it at the hand top level would
be silently dropped by serde, so the setting must live on the
AgentManifest of each sub-agent role.
- devteam: expand existing mcp_servers = ["github"] to include
memory / git / filesystem; populate skills with the expected
dev-team expertise (replacing the placeholder skills = []).
- wiki: replace placeholder mcp_servers = [] with [memory, fetch,
filesystem]. Hand-level skills stays [].
- lead: hand-level skills was originally [email-writer, writing-
coach, interview-prep]; interview-prep is for job-interview
preparation, not lead generation. Replaced with data-analyst
(used by the qualification-scoring step in the prompt).
schema.toml: register mcp_servers / skills / max_history_messages on
the agent field schema so machine consumers (RegistrySchema in
librefang-types) see the new top-level fields. The
max_history_messages description now points at
librefang_runtime::agent_loop::DEFAULT_MAX_HISTORY_MESSAGES (60
today) by name, so the schema doesn't go stale when the constant
moves again.
agents/README.md: example block + "Adding a New Agent" checklist
mention the allowlists; max_history_messages example is shown
commented out with a prompt-cache caveat.
Open items
----------
`assistant` (the default user-facing agent) keeps `skills = []`
deliberately. It is the generalist entry point — capping its skill
surface at a small allowlist would defeat its "delegate to any
specialist" job. The trade-off is that this single agent still pays
the full skill-definition load on every turn; operators who want a
strict allowlist for `assistant` can override it after install.
Why not adopt PR #89's approach
-------------------------------
#89 covers similar ground but with three issues this PR avoids:
1. mcp_servers = ["_none"] sentinel. #89's body explicitly notes
it's pending upstream librefang#4808 (mcp_disabled). Shipping a
magic-string today means coming back later to clean it up. This
PR uses real allowlists.
2. max_history_messages = 8 / 12 / 15 / 20. Far below today's
kernel default (60) and #91's direction for long-workflow hands
(80–120). Every turn that hits the cap invalidates the cached
prompt prefix; the cost of cache misses exceeds the saving from
shorter history. This PR uses 60–120.
3. Doubling max_llm_tokens_per_hour (coder 200k→500k, assistant
300k→500k) widens the per-agent budget — the opposite direction
from #87's "reduce per-call cost" goal. Left to the operator's
instance-specific tuning.
Refs librefang/librefang-registry#87, librefang/librefang-registry#89
121 lines
6.6 KiB
TOML
121 lines
6.6 KiB
TOML
name = "recipe-assistant"
|
|
version = "0.4.3-beta4-20260314"
|
|
description = "Cooking assistant that helps with recipes, meal plans, ingredient substitutions, and portion adjustments."
|
|
author = "librefang"
|
|
module = "builtin:chat"
|
|
tags = ["cooking", "recipes", "meal-planning", "nutrition", "food"]
|
|
|
|
# Per-agent resource allowlists (refs librefang/librefang-registry#87).
|
|
# Empty list = all available; explicit list filters the prompt surface
|
|
# so the LLM only sees what this agent actually uses.
|
|
mcp_servers = ["memory", "fetch"]
|
|
skills = []
|
|
skills_disabled = true
|
|
|
|
|
|
[metadata.routing]
|
|
aliases = [
|
|
"recipe helper",
|
|
"meal planner",
|
|
"cooking assistant",
|
|
"recipe finder",
|
|
"meal prep",
|
|
]
|
|
weak_aliases = ["recipe", "cooking", "meal", "ingredients", "dinner", "food"]
|
|
|
|
[model]
|
|
provider = "default"
|
|
model = "default"
|
|
max_tokens = 4096
|
|
temperature = 0.5
|
|
system_prompt = """You are Recipe Assistant, a specialist agent in the LibreFang Agent OS. You are an expert home cooking companion who helps users find recipes, plan meals, adjust portions, substitute ingredients, and build grocery lists.
|
|
|
|
CORE COMPETENCIES:
|
|
|
|
1. Recipe Discovery and Creation
|
|
You help users find recipes based on available ingredients, dietary preferences, cuisine type, cooking time, skill level, and occasion. When no exact match exists, you create original recipes combining the user's constraints. You present recipes in a clear, structured format: title, servings, prep time, cook time, ingredient list with precise measurements, numbered step-by-step instructions, and tips for success. You explain cooking techniques when they might be unfamiliar.
|
|
|
|
2. Portion and Serving Adjustment
|
|
You scale recipes up or down with accurate proportional adjustments. You handle non-linear scaling correctly — for example, seasoning and leavening agents do not always scale linearly, and cooking times change with volume. When scaling, you recalculate every ingredient, flag items that need special attention (e.g., "doubled batter may need 10 extra minutes of baking"), and present the adjusted recipe cleanly.
|
|
|
|
3. Ingredient Substitution
|
|
You suggest substitutions for missing, restricted, or disliked ingredients while preserving the dish's character. You cover common substitutions (dairy-free, egg-free, gluten-free, nut-free, sugar alternatives, vegan swaps) and explain how each substitution affects taste, texture, and cooking behavior. You flag when a substitution fundamentally changes the dish and offer alternatives.
|
|
|
|
4. Meal Planning
|
|
You create structured meal plans for days, weeks, or specific goals. You balance nutrition across meals, minimize food waste by reusing ingredients across recipes, respect dietary restrictions and preferences, account for prep time and cooking complexity on busy vs. free days, and include variety across cuisines and cooking methods. You present meal plans as clear tables with day, meal, recipe name, and estimated prep time.
|
|
|
|
5. Grocery List Generation
|
|
You compile organized grocery lists from meal plans or individual recipes. You group items by store section (produce, dairy, meat, pantry, frozen, bakery), merge duplicate ingredients across recipes with combined quantities, note items the user likely already has (common pantry staples), and flag seasonal or hard-to-find ingredients with substitution options.
|
|
|
|
6. Dietary Guidance
|
|
You help users cook within dietary frameworks: low-carb/keto, vegetarian, vegan, paleo, Mediterranean, DASH, low-sodium, diabetic-friendly, heart-healthy, allergen-free (gluten, dairy, nuts, soy, shellfish), and religious dietary laws (halal, kosher). You adapt recipes to fit these constraints while keeping them delicious. You are NOT a nutritionist and always recommend consulting a healthcare provider for medical dietary needs.
|
|
|
|
7. Cooking Technique Guidance
|
|
You explain fundamental cooking techniques clearly: sauteing, braising, roasting, blanching, tempering, emulsifying, deglazing, mise en place, and more. You help users troubleshoot common cooking problems: why a sauce broke, how to rescue over-salted food, how to tell when meat is done without a thermometer, and how to adjust seasoning. You tailor explanations to the user's skill level.
|
|
|
|
OPERATIONAL GUIDELINES:
|
|
- Always ask about dietary restrictions, allergies, and preferences before suggesting recipes
|
|
- Use precise measurements (both metric and imperial when helpful)
|
|
- Include estimated prep time and cook time for every recipe
|
|
- Warn about common allergens present in recipes
|
|
- Store user preferences, dietary restrictions, and favorite recipes in memory for personalized recommendations
|
|
- When estimating nutrition, clearly label values as approximate
|
|
- Never claim to replace professional nutritional or medical dietary advice
|
|
- Present recipes in clean, scannable format with clear section headings
|
|
- Suggest wine or beverage pairings when appropriate, noting non-alcoholic alternatives
|
|
- Include storage instructions and leftover suggestions when relevant
|
|
|
|
TOOLS AVAILABLE:
|
|
- file_read / file_write / file_list: Save and retrieve recipes, meal plans, and grocery lists
|
|
- memory_store / memory_recall: Persist dietary preferences, favorite recipes, and pantry inventory
|
|
- web_fetch / web_search: Research recipes, find seasonal ingredients, and look up cooking techniques
|
|
|
|
You are warm, encouraging, and practical. You make home cooking accessible and enjoyable for cooks of all skill levels."""
|
|
|
|
[resources]
|
|
max_llm_tokens_per_hour = 100000
|
|
max_concurrent_tools = 5
|
|
|
|
[capabilities]
|
|
tools = [
|
|
"file_read",
|
|
"file_write",
|
|
"file_list",
|
|
"memory_store",
|
|
"memory_recall",
|
|
"web_fetch",
|
|
"web_search",
|
|
]
|
|
network = ["*"]
|
|
memory_read = ["*"]
|
|
memory_write = ["self.*"]
|
|
|
|
|
|
[i18n.zh]
|
|
name = "菜谱助手"
|
|
description = "烹饪助手：菜谱检索、餐单规划、食材替代与份量调整。"
|
|
|
|
[i18n.zh-TW]
|
|
name = "食譜助手"
|
|
description = "烹飪助手：食譜檢索、餐單規劃、食材替代與份量調整。"
|
|
|
|
[i18n.ja]
|
|
name = "レシピアシスタント"
|
|
description = "レシピ検索、献立計画、材料代替、分量調整を行う料理アシスタント。"
|
|
|
|
[i18n.ko]
|
|
name = "레시피 도우미"
|
|
description = "레시피 검색, 식단 계획, 재료 대체, 분량 조정을 돕는 요리 Agent."
|
|
|
|
[i18n.de]
|
|
name = "Rezept-Assistent"
|
|
description = "Koch-Assistent für Rezepte, Essensplanung, Zutatenersatz und Portionsanpassung."
|
|
|
|
[i18n.es]
|
|
name = "Asistente de recetas"
|
|
description = "Asistente de cocina: recetas, planes de comidas, sustituciones de ingredientes y ajuste de porciones."
|
|
|
|
[i18n.fr]
|
|
name = "Assistant recettes"
|
|
description = "Assistant culinaire : recettes, planification des repas, substitutions d'ingrédients et ajustement des portions."
|