Curated Agent Skills scene packs for AI tools: 57 packs / 403 skills / 27 domains / 108 chains — one pack = one real-world scenario; drop-in for any SKILL.md-compatible tool.
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
feat(repo): 双主线显性化 + 专家团四平台安装包可下载(专家团升格为并列产品线) - 三语 README:新增「两条产品线——互不冲突」总览表;专家团升格为产品线 B 独立小节(定位 / 18 支团队清单 / 安装包下载表 / 来源说明),技能包保留为产品线 A - 官网首页新增双主线卡片(计数由 build_site.py 锚点注入,防静默漂移); 专家团子页新增下载区 + 返回技能库入口 - export-platforms.py 新增可复现 zip(opencode/claude/cursor/gemini/all)→ site/downloads/(GitCode Pages 从 main 的 /site 部署,必须入库) - CI 新增安装包摘要门禁:重建产物 ↔ site/downloads 逐包比对,防漂移 - 门禁:validate 0/0、pytest 52 passed、子树 verify + 19 单测 + site --check、 下载包摘要一致、py_compile 全仓无语法错误 | 12 小时前 | |
fix(audit): 全仓查漏补缺——293 处相对引用修复、上游资产补齐(CONNECTORS/HyperFrames 预设/workbench 指南)、gws 12→27、pack 字段归一、假 .json 改名(418 技能 / 112 链) | 38 分钟前 | |
feat(repo): 双主线显性化 + 专家团四平台安装包可下载(专家团升格为并列产品线) - 三语 README:新增「两条产品线——互不冲突」总览表;专家团升格为产品线 B 独立小节(定位 / 18 支团队清单 / 安装包下载表 / 来源说明),技能包保留为产品线 A - 官网首页新增双主线卡片(计数由 build_site.py 锚点注入,防静默漂移); 专家团子页新增下载区 + 返回技能库入口 - export-platforms.py 新增可复现 zip(opencode/claude/cursor/gemini/all)→ site/downloads/(GitCode Pages 从 main 的 /site 部署,必须入库) - CI 新增安装包摘要门禁:重建产物 ↔ site/downloads 逐包比对,防漂移 - 门禁:validate 0/0、pytest 52 passed、子树 verify + 19 单测 + site --check、 下载包摘要一致、py_compile 全仓无语法错误 | 12 小时前 | |
fix(audit): 全仓查漏补缺——293 处相对引用修复、上游资产补齐(CONNECTORS/HyperFrames 预设/workbench 指南)、gws 12→27、pack 字段归一、假 .json 改名(418 技能 / 112 链) | 38 分钟前 | |
chore(release): v0.23.1 — 418 技能 / 57 包 / 27 域 / 112 链(全仓查漏补缺) | 34 分钟前 | |
fix(audit): 全仓查漏补缺——293 处相对引用修复、上游资产补齐(CONNECTORS/HyperFrames 预设/workbench 指南)、gws 12→27、pack 字段归一、假 .json 改名(418 技能 / 112 链) | 38 分钟前 | |
fix(skills): 批次5命令级可运行性清零 — 64技能定向扫描全PASS - 修复范式(打样验证后四组并行执行):危险/占位符/多行续行示例 # 注释化或改干跑;/path/to 与 <ID> 占位换随包真实样例;为 40+ 技能新增 assets/ 与 examples/ 可运行样例(样例数据按脚本数据类精确构造) - 脚本真bug修3处:migration_generator、upgrade_planner、e2e_scaffold(--out 为 .json 时文件进同名目录+摘要落精确路径) - runner real_scenario_test.py 修4处(场景构造与路径解析) - 新增 tools/campaign_inventory.py 战役盘点工具与产物;tests/_effectiveness/batch5_fixlists/ 四组修复清单入库 - 覆盖范围:g1 架构/规划10技能、g2 编程其余17技能、g3 writing/tools/office/integrations 21技能、g4 零散域13技能 + 打样4技能(重叠合并后64技能65条) - 门禁:real_scenario_test.py --with-llm 定向扫描 65/65 PASS;pytest 589 passed 1 skipped | 5 天前 | |
feat(search): 中文检索(description_zh 全仓覆盖)与 find_skill --group 场景库联动 - 为 166 个原仅英文描述的在库技能补 frontmatter description_zh(连同本批 新技能,全仓 339 技能均可中文路由;另 37 个技能原生中文描述) - 站点:技能卡片优先展示中文描述、搜索命中 desc_zh;app.js 增 CJK 宽松 匹配(「合同审查」可命中「审查合同」)并逐字高亮 - find_skill.py:搜索加 description_zh 权重与 CJK 字符集匹配;新增 search --group <场景库>(id/中文名/英文名,读 taxonomy.json) - build_site.py:技能卡片输出 desc_zh 字段 | 2 小时前 | |
chore: 补充项目规范文件(.editorconfig、.gitattributes) | 4 天前 | |
chore: 补充项目规范文件(.editorconfig、.gitattributes) | 4 天前 | |
feat(writing/video/design/education): 配方v2第四批 — 主观创造五技能带出处重写 + A/B盲评台架三轮实测 - 五技能 1.x→2.0 全文重写:article-outliner / video-script-writer / storyboard-designer / image-prompt-engineer / own-voice-rewrite - 方法论全部来自真实从业者来源(Minto/短视频钩子/镜头语法/生图五段/计量文体学),来源落盘 references/sources-and-methodology.md - 新增 tools/effectiveness_ab.py A/B 盲评台架 + effectiveness_tasks.json(rubric 锚定外部标准) - 三轮 45 次任务测量:台架抓出两类我方回归(可选输入阻塞交付、多风格分流改设计)并修复;最终 new 8 / old 2 / tie 4 / invalid 1,三轮合计 new 22 / old 7 - 脚本修复:outliner 要点轮转+word_count_target+placeholders;script_writer 0秒场景+duration_note+caption_check - storyboard 随包 examples/ 两镜样例,文档示例开箱可运行 - 证据归档 tests/_effectiveness/(round1-3 + addendum + 全库命令级扫描),报告 batch4_evidence.md | 5 天前 | |
chore(site): 站点镜像与文档收尾——中文检索与 12 新包全量入站 - site/:site.json 与 site/skills 镜像按当前 339 技能重建,新增 12 包 zip 入站 - 三语 README / AGENTS / SOURCES / docs/TAXONOMY-V2 与 manifest.json 同步 | 2 小时前 | |
fix(audit): 全仓查漏补缺——293 处相对引用修复、上游资产补齐(CONNECTORS/HyperFrames 预设/workbench 指南)、gws 12→27、pack 字段归一、假 .json 改名(418 技能 / 112 链) | 38 分钟前 | |
Initial commit: SkillKit curated skill packs | 1 个月前 | |
fix(audit): repair P0/P1 defects + reconcile docs to 163/39 (v0.22) Scripts: drop dead publish_common import (generate_cover inlines stdlib HTTP + --execute gate); recursive batch discovery in quality_scorer/security_scorer with 0-script null scoring; pii_scan default dry-run (write requires --execute, SKILL.md synced); CJK sentence splitting + real char_count in content-editor; stable sha256 cache key + atomic session write in pipeline; null-skip grounding + score renormalization in eval_harness; correct web-search path in research_agent. Tests: 7 new regression suites (34 passed total). Docs: SOURCES.md reconciled 154->163 = upstream 32 + adapted 5 + self 126; add 14 missing self-built skills, drop 5 deleted skill rows and dead _common helper row; CONTRIBUTING drops _common guidance; README badges 156->163 (zh/en/ja); site/index.html counts synced to site.json. Chore: remove committed deep-research error artifact research_677e03a8.json. | 1 天前 | |
chore: unify copyright and drop old identity refs | 22 天前 | |
fix(audit): 全仓查漏补缺——293 处相对引用修复、上游资产补齐(CONNECTORS/HyperFrames 预设/workbench 指南)、gws 12→27、pack 字段归一、假 .json 改名(418 技能 / 112 链) | 38 分钟前 | |
fix(audit): 全仓查漏补缺——293 处相对引用修复、上游资产补齐(CONNECTORS/HyperFrames 预设/workbench 指南)、gws 12→27、pack 字段归一、假 .json 改名(418 技能 / 112 链) | 38 分钟前 | |
fix(audit): 全仓查漏补缺——293 处相对引用修复、上游资产补齐(CONNECTORS/HyperFrames 预设/workbench 指南)、gws 12→27、pack 字段归一、假 .json 改名(418 技能 / 112 链) | 38 分钟前 | |
Initial commit: SkillKit curated skill packs | 1 个月前 | |
fix(audit): 全仓查漏补缺——293 处相对引用修复、上游资产补齐(CONNECTORS/HyperFrames 预设/workbench 指南)、gws 12→27、pack 字段归一、假 .json 改名(418 技能 / 112 链) | 38 分钟前 | |
fix(manifest): sync skills from pack.json for existing packs; site skill-count clarity - build.py sync_manifest: existing entries never refreshed their skills list, so ai-research-writing silently dropped 12 paper-domain skills (index showed 100 vs 112 in packs/*/pack.json). Now skills is always upserted from pack.json. - build_site.py: emit n_skills_on_disk (113) alongside n_skills (112 packed); site hero shows 113 with a note that 112 are packed and 1 is a template. - CHANGELOG: correct badge count to 113 (112 packed + 1 sample-skill template). Gate: pytest 257 passed/1 skipped, validate_skills 111 skills 0 errors 0 warnings. | 11 天前 | |
feat(hub): skill validator, release helper, metadata migration; flat installer; sha256 in manifest | 1 个月前 | |
fix(project): kill dual-build drift, un-CI the fake-green test runner, declare deps Project-level issue sweep findings and fixes: - [P1 high] build.sh excluded-patterns failed on nested paths (__pycache__/*, .gitignore) — its _all.zip shipped 240 junk files; tri-lingual READMEs still recommended it. Deleted; build.py is now the single build entry. - [P2 high] CI tests job ran test_*.py via bare subprocess — pytest-style assert functions never execute (exit 0 = fake green). Replaced with python -m pytest skills -q, matching CONTRIBUTING.md. - [P3 med] No Python dependency declaration for bundled scripts. Added requirements.txt grouped by purpose (scripts degrade gracefully without it). - [P4 low] tools/: 3 one-shot migration scripts (432 lines, zero references) removed. - [P5 low] CHANGELOG: added Keep-a-Changelog compare-links section (15 tags). - [P6/P7 low] pytest.ini added; PytestCollectionWarning 3 -> 0. Cleared as non-issues: git history has no real secrets (ghp_FAKE... is the env-secrets-manager teaching fixture); zero cross-skill deps; zero doc dead links; TODO/GitHub references are legit content. Regression: gate PASSED, pytest 223 passed / 0 warnings, 22 archives built. | 13 天前 | |
chore(release): v0.23.1 — 418 技能 / 57 包 / 27 域 / 112 链(全仓查漏补缺) | 34 分钟前 | |
docs(release): 数字对账 + SOURCES 三线归属补登记 + dist/site 回填 - 三语 README: 徽章 skills-143→154, chains 58→62, 饼图改三线口径 116/33/5(154) - SOURCES.md: 137/34→154/36, 两条线→三条线(上游精选33+上游改造5+自建116), 补登记 v0.16 以来未入表的 33 个自研技能与 5 个上游改造技能, 追加 2026-09-21 历史条目 - docs/FULL-TEST-REPORT.md: 新增第八节「场景级真实验收」+ 复现命令更新(9节) - pytest.ini: --import-mode=importlib 修复 93 个同名 test_smoke_all.py 全树收集撞名 (80 errors → 566 passed, 1 skipped) - manifest.json: build.py 回填 size_kb/sha256(20 个脚本改动后 dist 漂移修正) - site/: packs zip 全量重建, site.json 同步 | 6 天前 | |
ci(deps): pin CI test deps + requirements to 2026-09-26 latest, gate pages deploy on validator, py3.13 (Gate A advisory open) | 1 天前 | |
feat(packs): P1 收尾 + P2 四源精选——5 个新场景包 / 64 技能(路线图结项) - caveman-toolkit(caveman 技能面 7,MIT;引擎目录 BSL-1.1 未收录)→ meta 域 - cybersecurity-pro(mukul975/Anthropic-Cybersecurity-Skills 818 挑 14,Apache-2.0)→ programming/security - gtm-growth(gooseworks-ai/goose-skills 281 挑 16,MIT)→ marketing/gtm - language-standards(mindrally/skills 240+ 挑 14,Apache-2.0)→ programming/standards - company-playbooks(cbrock84/headcount 172 挑 13,MIT)→ ops/leadership/finance/product P2 许可核验裁决(不迁):Lawvable(CC BY-NC-ND 4.0)/ OpenClaw Medical(无 LICENSE)/ buildwithclaude(聚合平台,授权链路混杂)。 仓库总量 403 技能 / 57 包 / 27 域 / 108 链(原 339 / 52 / 27 / 88);税务/链图谱/三语 README/ SOURCES/CHANGELOG/TAXONOMY-V2 同步;站点与安装包重建;门禁四连全绿(0 错 0 警 / pytest 284 过)。 | 2 小时前 |
加载中
项目介绍
Curated Agent Skills scene packs for AI tools: 57 packs / 403 skills / 27 domains / 108 chains — one pack = one real-world scenario; drop-in for any SKILL.md-compatible tool.
https://x33834.github.io/awesome-skillkit/定制我的领域