frostbyte_neo/yao-meta-skill
1
0
Fork 0
Code Issues Pull Requests Actions Deployments Models Agent Notes Packages Projects Releases Wiki Activity

Improve yao meta skill trigger coverage

Browse Source
This commit is contained in:
yaojingang
2026-07-02 11:29:19 +08:00
parent ab2b0e9fb5
commit cf70bbe6c2
4 changed files with 142 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
SKILL.md
+1 -1
View File
@@ -1,6 +1,6 @@
---
name: yao-meta-skill
description: Create, refactor, evaluate, and package agent skills from workflows, prompts, transcripts, docs, or notes. Use for skill creation, reusable workflow packaging, skill improvement, evals, and team-ready distribution.
description: Create, refactor, evaluate, package, migrate, govern, and release agent skills from workflows, prompts, transcripts, docs, notes, SOPs, scripts, or repeated team practices. Use when asked to create/build a skill, turn a repeated process into a reusable agent capability, improve or migrate an existing skill, optimize trigger routing, add evals/tests, add references/scripts/interface/manifest, prepare packaging/installation/release, or make a skill team-ready. Also trigger on Chinese asks such as 做/改/重构/迁移/发布 skill, 把流程/SOP/提示词/对话记录沉淀成可复用能力, 封装成团队可复用的 skill, 优化已有 skill, 补 trigger 评测, 收紧触发边界, or 做打包发布检查. Do not use for summary-only, translation-only, brainstorming-only, or documentation-only requests that explicitly say no skill or agent execution.
metadata:
author: Yao Team
---
evals/improved_description.txt
+1 -1
View File
@@ -1 +1 @@
Create, refactor, evaluate, and package agent skills from workflows, prompts, transcripts, docs, or notes. Use when asked to create a skill, turn a repeated process into a reusable skill, improve an existing skill, add evals, or package a skill for team reuse. Also trigger on common Chinese asks such as 做一个 skill, 把流程沉淀成可复用能力, 封装成团队可复用的 skill, 优化已有 skill, 补 trigger 评测, or 收紧触发边界.
Create, refactor, evaluate, package, migrate, govern, and release agent skills from workflows, prompts, transcripts, docs, notes, SOPs, scripts, or repeated team practices. Use when asked to create/build a skill, turn a repeated process into a reusable agent capability, improve or migrate an existing skill, optimize trigger routing, add evals/tests, add references/scripts/interface/manifest, prepare packaging/installation/release, or make a skill team-ready. Also trigger on Chinese asks such as 做/改/重构/迁移/发布 skill, 把流程/SOP/提示词/对话记录沉淀成可复用能力, 封装成团队可复用的 skill, 优化已有 skill, 补 trigger 评测, 收紧触发边界, or 做打包发布检查. Do not use for summary-only, translation-only, brainstorming-only, or documentation-only requests that explicitly say no skill or agent execution.
evals/semantic_config.json
+88 -1
View File
@@ -18,12 +18,19 @@
"turn a repeated process into a reusable skill",
"improve an existing skill",
"add evals",
"add tests",
"package a skill for team reuse",
"migrate an existing skill",
"prepare packaging, installation, or release",
"govern a reusable skill",
"做一个 skill",
"把流程沉淀成可复用 skill",
"优化现有 skill",
"补 trigger 评测",
"封装成团队可复用 skill"
"封装成团队可复用 skill",
"迁移 skill",
"发布 skill",
"打包发布检查"
],
"exclusions": [
"summary-only requests",
@@ -75,16 +82,29 @@
"formalize this process into a reusable capability",
"turn this playbook into an agent capability package",
"convert this into a maintained skill",
"agent capability",
"reusable agent capability",
"skill library entry",
"skill authoring",
"做一个 skill",
"创建一个 skill",
"新建一个 skill",
"帮我做个 skill",
"改一个 skill",
"重构一个 skill",
"迁移一个 skill",
"发布一个 skill",
"这个 skill",
"已有 skill",
"现有 skill",
"做成一个 skill",
"封装成一个 skill",
"沉淀成一个 skill",
"整理成一个 skill",
"抽成一个 skill",
"可复用 skill",
"可执行 skill",
"团队 skill",
"agent 能力包",
"能力包",
"skill 包"
@@ -103,19 +123,27 @@
"existing skill draft",
"rough notes",
"standard operating procedure",
"operating procedure",
"sop",
"playbook",
"transcript",
"chat transcript",
"meeting notes",
"workflow fragments",
"release notes",
"onboarding flow",
"support escalation routine",
"quarterly release routine",
"scripts",
"repeated team practice",
"流程",
"工作流",
"操作流程",
"做事流程",
"作业流程",
"方法",
"经验",
"复盘",
"sop 文档",
"会议纪要",
"对话记录",
@@ -142,14 +170,32 @@
"ops library",
"maintained",
"operationalize",
"release-ready",
"publish",
"release",
"installation",
"installable",
"portability",
"registry",
"governance",
"review gate",
"可复用",
"复用",
"封装",
"打包",
"沉淀",
"标准化",
"标准化下来",
"团队复用",
"给团队用",
"团队可用",
"安装复用",
"可安装",
"发布",
"上架",
"交付",
"治理",
"注册表",
"方法包",
"团队能力包",
"沉淀下来",
@@ -171,16 +217,32 @@
"false negatives",
"stress test",
"route boundary",
"routing",
"route evals",
"trigger routing",
"activation",
"description optimization",
"tests",
"install checks",
"release checks",
"hardening",
"评测",
"补评测",
"补 eval",
"补测试",
"触发测试",
"触发词",
"触发范围",
"召回",
"收紧触发边界",
"路由边界",
"路由评测",
"激活",
"误触发",
"漏触发",
"质量门",
"安装验证",
"发布检查",
"打包检查",
"边界测试"
]
@@ -198,11 +260,32 @@
"before sharing it with the team",
"tighten the boundary",
"revise this skill",
"refactor this skill",
"migrate this skill",
"port this skill",
"convert this skill",
"release this skill",
"publish this skill",
"make this skill release-ready",
"codex-ready package",
"claude skill",
"codex format",
"interface metadata",
"manifest",
"install checks",
"draft skill",
"harden this skill for library reuse",
"优化这个 skill",
"改进这个 skill",
"重构这个 skill",
"扩大这个 skill",
"调整这个 skill",
"优化触发范围",
"扩大触发范围",
"迁移这个 skill",
"发布这个 skill",
"把这个 skill 迁移",
"让这个 skill 可发布",
"补一下评测",
"补 trigger 测试",
"加强路由判断",
@@ -351,12 +434,16 @@
"do not convert it into a reusable capability",
"keep it as documentation only",
"discussion only",
"no agent execution",
"先不要做成 skill",
"先不做 skill",
"不要做成 skill",
"不用封装",
"不要打包",
"不要沉淀成能力",
"先别做成可复用能力",
"不需要 agent 执行",
"不做 agent 能力",
"只做文档",
"只做说明"
]
evals/trigger_cases.json
+52
View File
@@ -54,6 +54,18 @@
"text": "Add trigger evals to this skill before sharing it with the team.",
"family": "iterate_existing_skill"
},
{
"text": "Migrate this Claude skill into a Codex-ready package with interface metadata and install checks.",
"family": "migrate_existing_skill"
},
{
"text": "Make this existing skill release-ready: add manifest/interface metadata, packaging checks, and trigger evals.",
"family": "release_packaging"
},
{
"text": "Optimize the activation wording for this skill so broader skill-authoring requests are routed correctly.",
"family": "trigger_expansion"
},
{
"text": "We have a messy release runbook, export process, and a prompt history; turn all of that into one reusable skill package with evals and packaging checks.",
"family": "complex_multi_asset"
@@ -133,6 +145,18 @@
{
"text": "把这个客服升级流程标准化成一个可执行的 skill，并加上边界测试。",
"family": "cn_eval_and_package"
},
{
"text": "把这个 Claude skill 迁移成 Codex 可用的 skill 包，补 interface、manifest 和安装验证。",
"family": "cn_migrate_existing_skill"
},
{
"text": "把这套提示词和脚本整理成团队能安装复用的 skill，并做打包发布检查。",
"family": "cn_release_packaging"
},
{
"text": "扩大这个 skill 的触发范围，同时补路由评测，避免漏触发。",
"family": "cn_trigger_expansion"
}
],
"should_not_trigger": [
@@ -239,6 +263,22 @@
{
"text": "先一起头脑风暴几个方向，不要封装成 skill。",
"family": "cn_brainstorm_only"
},
{
"text": "Polish this release note for publishing, but do not do any skill work.",
"family": "document_only"
},
{
"text": "Migrate this wiki article into a new docs folder; do not create an agent skill.",
"family": "document_only"
},
{
"text": "把这篇 SOP 改成知识库文档，不需要 agent 执行。",
"family": "cn_document_only"
},
{
"text": "检查安装步骤是否清楚，只做 README，不做 skill 包。",
"family": "cn_document_only"
}
],
"near_neighbor": [
@@ -333,6 +373,18 @@
{
"text": "这次只给我一个一次性 prompt，不用复用成 skill。",
"family": "cn_one_off_vs_reusable"
},
{
"text": "Create a reusable onboarding template for humans, not an agent skill.",
"family": "document_export_vs_agent_skill"
},
{
"text": "Draft a release checklist for a future skill, but do not package or evaluate the skill yet.",
"family": "future_outline_vs_build"
},
{
"text": "整理成团队发布流程文档即可，不要做成 skill，也不要做触发评测。",
"family": "cn_document_export_vs_agent_skill"
}
]
}