Merge remote-tracking branch 'origin/stack/agent-profiles-1-seam' into stack/agent-profiles-3-wire
# Conflicts: # docs/module-graph.md # packages/host/apiproxy/src/api-proxy.ts # packages/host/apiproxy/tsconfig.json
This commit is contained in:
Yichen Jiang
@@ -2,5 +2,5 @@
|
||||
# side as of the last confirmed-consistent state. Both languages carry equal authority;
|
||||
# after editing either side, bring the other along and re-record with:
|
||||
# pnpm run verify-translation-pairing --write packages/skill/tool-skill/README.md
|
||||
README.md: 5d7747593d910c38f8008345c3cea3d991dc5daf
|
||||
README.zh.md: 844f204c43f7cbed99aff40f03a6086868550f36
|
||||
README.md: 704eb7eb1f611c20f76ce79190326296ff63da42
|
||||
README.zh.md: 09fb5b954ea595b962f7670ce09db1ae22dce29c
|
||||
|
||||
@@ -36,7 +36,7 @@ Tool execution does not add a synthetic context message. Its freshly loaded resu
|
||||
|
||||
#### What the model sees
|
||||
|
||||
If model-invocable skills exist and this exact `skill` tool is visible, the agent receives the catalog template below as a durable user-role message before the first request, with one data-dependent entry per sorted skill. Later membership, description, or visibility changes append a complete replacement using the same `<available_skills>` envelope; deleting every skill appends an empty envelope with an explicit instruction not to use older names. The template's closing sentence is the seam rule against double-loading: the user-explicit gesture boundary (the pre-step listener below) injects the same `renderSkillContent` output (shared from `@deepseek-ai/dsh-skill`) inline, and the catalog tells the model to follow that block instead of re-loading the skill through the tool; the replacement-catalog template carries the same sentence in both arms, including the emptied catalog.
|
||||
If model-invocable skills exist and this exact `skill` tool is visible, the agent receives the catalog template below as a durable user-role message before the first request, with one data-dependent entry per sorted skill. Later membership, description, or visibility changes append a complete replacement using the same `<available_skills>` envelope; deleting every skill appends an empty envelope with an explicit instruction not to use older names. The template's closing sentence is the rule against double-loading: the user-explicit gesture boundary (the pre-step listener below) injects the same `renderSkillContent` output (shared from `@deepseek-ai/dsh-skill`) inline, and the catalog tells the model to follow that block instead of re-loading the skill through the tool; the replacement-catalog template carries the same sentence in both arms, including the emptied catalog.
|
||||
|
||||
##### Skill catalog template
|
||||
|
||||
|
||||
@@ -36,7 +36,7 @@
|
||||
|
||||
#### 模型看到的内容
|
||||
|
||||
如果存在模型可调用 skill，且可见的正是这个 `skill` 工具，agent 会在第一个请求之前收到下方目录模板，其中包含每个已排序 skill 的一条随数据而定的条目。该目录是一条持久的用户角色消息。后续成员关系、描述或可见性的变化会使用同一个 `<available_skills>` 信封追加完整替换；删除所有 skill 时，会追加一个空信封，并明确指示不得使用旧名称。模板的结尾一句是防止双重加载的 seam 规则：用户显式的手势边界（下文的 pre-step 监听器）会把同一份 `renderSkillContent` 输出（共享自 `@deepseek-ai/dsh-skill`）内联注入，目录则告诉模型遵循该块，而不是再经工具重新加载该 skill；替换目录模板的两个分支——包括清空后的目录——都携带同一句话。
|
||||
如果存在模型可调用 skill，且可见的正是这个 `skill` 工具，agent 会在第一个请求之前收到下方目录模板，其中包含每个已排序 skill 的一条随数据而定的条目。该目录是一条持久的用户角色消息。后续成员关系、描述或可见性的变化会使用同一个 `<available_skills>` 信封追加完整替换；删除所有 skill 时，会追加一个空信封，并明确指示不得使用旧名称。模板的结尾一句是防止双重加载的规则：用户显式的手势边界（下文的 pre-step 监听器）会把同一份 `renderSkillContent` 输出（共享自 `@deepseek-ai/dsh-skill`）内联注入，目录则告诉模型遵循该块，而不是再经工具重新加载该 skill；替换目录模板的两个臂——包括清空后的目录——都携带同一句话。
|
||||
|
||||
##### Skill 目录模板
|
||||
|
||||
@@ -149,11 +149,11 @@ Load referenced resources only as needed.
|
||||
|
||||
#### 模型看到的内容
|
||||
|
||||
已认领的用户消息中任意位置、以空白为界、指名工作区目录中某个用户可调用 skill 的 `/name` token，会把该 skill 的完整 `<skill_content>` 渲染（与上文结果模板完全相同的形态）作为 `user` 角色的指令上下文注入，追加在该步骤所有其他注入之后——背景在前，模型要着手处理的材料在最后。只扫描直接的用户输入，检查在已加载定义上进行，未知名称和用户不可调用的名称保持为普通行文。这是 `disable-model-invocation` skill 唯一的入口，目录和 `skill` 工具永不暴露这类 skill；目录的结尾一句会告诉模型遵循注入块，而不是重新加载它。
|
||||
已认领用户消息中任意位置、以空白为界、指名工作区目录中某个用户可调用 skill 的 `/name` token，会把该 skill 的完整 `<skill_content>` 渲染（与上文结果模板完全相同的形态）作为 `user` 角色的指令上下文注入，追加在该步骤所有其他注入之后——背景在前，模型要着手处理的材料在最后。只扫描直接的用户输入，检查在已加载定义上进行，未知名称和用户不可调用的名称保持为普通行文。这是 `disable-model-invocation` skill 唯一的入口，目录和 `skill` 工具永不暴露这类 skill；目录的结尾一句会告诉模型遵循注入块，而不是重新加载它。
|
||||
|
||||
#### Token 影响
|
||||
|
||||
每次手势会把一份渲染后的 skill 正文作为注入上下文加进该轮次——其大小与同一 skill 的工具结果相同；这项成本会在用户请求时确定产生，而非由模型自行裁量。同一步骤内对同一 skill 的重复手势只注入一次。
|
||||
每次手势会把一份渲染后的 skill 正文作为注入上下文加进该轮次——尺寸与同一 skill 的工具结果相同，按用户的请求确定性地支付，而非由模型自行裁量。同一步骤内对同一 skill 的重复手势只注入一次。
|
||||
|
||||
#### KV Cache 影响
|
||||
|
||||
|
||||
@@ -26,7 +26,7 @@ export const inject = ['agents', 'tools', 'skills']
|
||||
|
||||
const DEFAULT_CATALOG_DESCRIPTION_MAX_LENGTH = 500
|
||||
/**
|
||||
* Durable provenance for one published session skill catalog. The catalog is a
|
||||
* Durable provider and item records for one published session skill catalog. The catalog is a
|
||||
* `catalog`-form context, so it records the entries it published beside the
|
||||
* model-facing prose: a consumer presenting the list must not re-parse the
|
||||
* `<available_skills>` block, whose framing exists for the model.
|
||||
|
||||
Reference in New Issue
Block a user