version: "1.0.1" name: agent-knowledge description: Retrieve from knowledge bases explicitly configured on the current Agent. when_to_use: Use this skill when a published or agent-scoped assistant needs context from its configured knowledge bases. provider_type: builtin provider_id: knowledge runtime_type: tool tools:

• retrieve_agent_knowledge

max_calls_per_turn: 20 timeout_seconds: 30 tool_governance: retrieve_agent_knowledge: tool_id: knowledge.retrieve_agent skill_id: agent-knowledge domain: knowledge effect: read asset_type: knowledge_base risk_level: low requires_asset_resolution: false reversible: false bulk_sensitive: false external_side_effect: false permission_scopes:

• knowledge:read

default_approval_policy: auto_by_permission_tier allowed_permission_tiers:

• basic
• advanced
• full

audit_required: true idempotency_required: false display: icon: book-open-check category: knowledge_retrieval scenarios:

• knowledge_research
• technical_development

label: en_US: Agent Knowledge zh_Hans: 智能体知识库 description: en_US: Designed for Agent answers that need internal sources; retrieves relevant facts and context only from knowledge bases bound to the current Agent. zh_Hans: 适用于智能体基于内部资料回答问题，可从当前智能体绑定的知识库中检索相关事实和来源上下文。 when_to_use: en_US: Use for Agent answers that need configured knowledge base retrieval. zh_Hans: 当智能体回复需要检索已绑定知识库时使用。 tags: en_US:

• Knowledge
• Agent

zh_Hans:

• 知识库
• 智能体

Agent Knowledge Skill

Use this skill to retrieve context only from knowledge bases configured on the current Agent. Knowledge access is authorized when the editor binds the knowledge bases to the Agent, and runtime retrieval uses that Agent binding grant.

Workflow

1. Call retrieve_agent_knowledge when the answer depends on the Agent's configured knowledge, product facts, policy, documentation, or other long-lived source material.
2. Use a concise query derived from the user's intent. Do not enumerate all user-accessible knowledge bases.
3. Do not ask for, guess, or pass dataset IDs. The backend reads configured knowledge bases from the Agent config.
4. Inspect status, source_summary, context_blocks, and scores before answering:

• If status is success and the retrieved blocks answer the user, answer from those blocks.
• If results are missing, weak, or off-topic, rewrite the query once using clearer entities, synonyms, or constraints from the user question and retry.
• After two retrieval attempts, if the relevant answer is still unclear, ask the user for clarification or say that no configured relevant knowledge was found.

1. When the user asks for original wording, definitions, synopsis text, policy clauses, exact wording, or "what does it say", quote or closely excerpt the retrieved source text first and cite the source. Summarize only when the user asks for a summary or the original text is too long.
2. When retrieved context is used, cite source names from source_summary or retriever_resources when useful.
3. Never expose internal IDs such as dataset ID, document ID, or segment ID to the user.
4. Never invent a knowledge-base answer when no relevant configured context was found.

Tool Usage

retrieve_agent_knowledge accepts:

• query: the user question or refined search query.
• top_k: optional maximum retrieved chunks. Defaults to 5 and is capped at 20.
• retrieval_mode: optional hybrid, vector, or graph. Omit it for the default hybrid mode; use graph only for relationship or entity questions.