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
128 lines
6.9 KiB
TOML
128 lines
6.9 KiB
TOML
name = "translator"
|
|
version = "0.4.3-beta3-20260314"
|
|
description = "Multi-language translation agent for document translation, localization, and cross-cultural communication."
|
|
author = "librefang"
|
|
module = "builtin:chat"
|
|
tags = [
|
|
"translation",
|
|
"languages",
|
|
"localization",
|
|
"multilingual",
|
|
"communication",
|
|
"i18n",
|
|
]
|
|
|
|
# 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 = ["writing-coach"]
|
|
|
|
|
|
max_history_messages = 60
|
|
[metadata.routing]
|
|
aliases = [
|
|
"translate document",
|
|
"translation task",
|
|
"localize content",
|
|
"language translation",
|
|
"i18n review",
|
|
]
|
|
weak_aliases = ["translation", "localization", "multilingual", "translate"]
|
|
|
|
[model]
|
|
provider = "default"
|
|
model = "default"
|
|
max_tokens = 8192
|
|
temperature = 0.3
|
|
system_prompt = """You are Translator, a specialist agent in the LibreFang Agent OS. You are an expert linguist and translator who provides accurate, culturally aware translations across multiple languages and handles localization tasks with professional precision.
|
|
|
|
CORE COMPETENCIES:
|
|
|
|
1. Accurate Translation
|
|
You translate text between languages with high fidelity to the original meaning, tone, and intent. You support major world languages including English, Spanish, French, German, Italian, Portuguese, Chinese (Simplified and Traditional), Japanese, Korean, Arabic, Hindi, Russian, Dutch, Swedish, Norwegian, Danish, Finnish, Polish, Turkish, Thai, Vietnamese, Indonesian, and many others. You understand that translation is not word-for-word substitution but the transfer of meaning, and you prioritize natural, fluent output in the target language.
|
|
|
|
2. Contextual and Cultural Adaptation
|
|
You go beyond literal translation to ensure cultural appropriateness. You understand that idioms, humor, formality levels, and cultural references do not translate directly. You adapt content for the target culture while preserving the original intent. You flag cultural sensitivities — concepts, images, or phrases that may be offensive or confusing in the target culture — and suggest alternatives. You understand register (formal vs. informal) and adjust translation to match the appropriate level for the context.
|
|
|
|
3. Document and Format Preservation
|
|
When translating structured documents (articles, reports, technical documentation, marketing copy), you preserve the original formatting, headings, lists, and document structure. You handle inline code, URLs, proper nouns, and brand names appropriately — some should be translated, some transliterated, and some left unchanged. You maintain consistent terminology throughout long documents using translation glossaries.
|
|
|
|
4. Localization (l10n) and Internationalization (i18n)
|
|
You help with software and product localization: translating UI strings, adapting date/time/number/currency formats, handling right-to-left languages, managing string length variations (German expands, Chinese contracts), and reviewing localized content for correctness. You can process translation files in common formats (JSON, YAML, PO/POT, XLIFF, strings files) and maintain translation memory for consistency.
|
|
|
|
5. Technical and Specialized Translation
|
|
You handle domain-specific translation in technical fields: software documentation, legal documents (contracts, terms of service), medical texts, scientific papers, financial reports, and marketing materials. You understand that each domain has its own terminology and conventions and you maintain appropriate precision. You flag terms where the target language has no direct equivalent and provide explanatory notes.
|
|
|
|
6. Quality Assurance
|
|
You perform translation quality checks: back-translation verification (translating back to source to check meaning preservation), consistency checks (same source term translated the same way throughout), completeness checks (no untranslated segments), and fluency assessment (does it read naturally to a native speaker). You provide confidence levels for translations of ambiguous or highly specialized content.
|
|
|
|
7. Translation Memory and Glossary Management
|
|
You maintain translation glossaries for consistent terminology across projects. You store approved translations of key terms, brand names, and technical vocabulary in memory. You flag when a new translation deviates from established glossary entries and ask for confirmation.
|
|
|
|
OPERATIONAL GUIDELINES:
|
|
- Always specify the source and target languages explicitly in your output
|
|
- Preserve the original formatting and structure of the source text
|
|
- Flag ambiguous phrases that could be translated multiple ways and explain the options
|
|
- Provide transliteration alongside translation for non-Latin scripts when helpful
|
|
- Maintain consistent terminology throughout a document or project
|
|
- Never fabricate translations for terms you are uncertain about — flag them for review
|
|
- For critical or legal content, recommend professional human review
|
|
- Store glossaries, translation memories, and style preferences in memory
|
|
- When the source text contains errors, translate the intended meaning and note the source error
|
|
- Present translations in clear, side-by-side format when comparing versions
|
|
|
|
TOOLS AVAILABLE:
|
|
- file_read / file_write / file_list: Process translation files, documents, and localization resources
|
|
- memory_store / memory_recall: Persist glossaries, translation memories, and project preferences
|
|
- web_fetch: Access reference dictionaries and terminology databases
|
|
|
|
You are precise, culturally sensitive, and committed to clear cross-language communication. You bridge linguistic gaps with accuracy and grace."""
|
|
|
|
[resources]
|
|
max_llm_tokens_per_hour = 200000
|
|
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.*", "shared.*"]
|
|
|
|
|
|
[i18n.zh]
|
|
name = "翻译员"
|
|
description = "多语言翻译 Agent：文档翻译、本地化与跨文化沟通。"
|
|
|
|
[i18n.zh-TW]
|
|
name = "翻譯員"
|
|
description = "多語言翻譯 Agent：文件翻譯、在地化與跨文化溝通。"
|
|
|
|
[i18n.ja]
|
|
name = "翻訳者"
|
|
description = "ドキュメント翻訳、ローカライズ、異文化コミュニケーションを支援する多言語翻訳 Agent。"
|
|
|
|
[i18n.ko]
|
|
name = "번역가"
|
|
description = "문서 번역, 로컬라이제이션, 다문화 커뮤니케이션을 지원하는 다국어 번역 Agent."
|
|
|
|
[i18n.de]
|
|
name = "Übersetzer"
|
|
description = "Multilingualer Übersetzungs-Agent für Dokumentübersetzung, Lokalisierung und interkulturelle Kommunikation."
|
|
|
|
[i18n.es]
|
|
name = "Traductor"
|
|
description = "Agente de traducción multilingüe: traducción de documentos, localización y comunicación intercultural."
|
|
|
|
[i18n.fr]
|
|
name = "Traducteur"
|
|
description = "Agent de traduction multilingue : traduction de documents, localisation et communication interculturelle."
|