[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7C43QAF92EH44JN5JKTP4JV":3,"skill-same:01M7C43QAF92EH44JN5JKTP4JV":54},{"skillNo":4,"name":5,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":6,"repositoryNo":11,"starsCount":12,"licenseSpdxId":13,"licenseFileUrl":6,"sourceHtmlUrl":14,"categories":15,"keywords":16,"publishedAt":17,"linkOnly":18,"markdownContent":19,"markdownBody":20,"frontMatter":21,"refName":22,"commitSha":23,"skillDirectory":24,"forksCount":12,"repositoryHtmlUrl":25,"repositoryHomepageUrl":25,"ownerHtmlUrl":26,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":27,"install":28,"resources":41},"01M7C43QAF92EH44JN5JKTP4JV","agent-accuracy-guard",null,"SillyHippy\u002Fzo-skills","SillyHippy","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F159052837?v=4","01M7C43A8PDYMCDPTNFYF2RSC1","01M7C43A9CBS76QPH4PPTKXX54",0,"MIT","https:\u002F\u002Fgithub.com\u002FSillyHippy\u002Fzo-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-accuracy-guard\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: agent-accuracy-guard\ndescription: Preflight guard for Zo\u002FHermes to prevent repeated failures: unauthorized actions, wrong provider\u002Fmodel assumptions, false completion claims, and ignoring Telegram delivery. Use before agent swarms, provider\u002Fmodel selection, sending files, document generation, or any high-risk action.\ncompatibility: Created for Zo Computer\nmetadata:\n  author: sillyhippy.zo.computer\n---\n# Agent Accuracy Guard\n\nUse this before high-risk actions.\n\n## Non-negotiable checks\n\n1. **Authorization**\n   - If the user asked for a plan first, stop after the plan.\n   - Do not run agent swarm, bulk jobs, paid APIs, email, or external sends without explicit approval.\n   - \"I guess continue\" counts as approval only if the user is clearly telling you to continue the same pending action.\n\n2. **Provider\u002Fmodel verification**\n   - Never infer model availability from a local proxy unless the user explicitly said to use that proxy.\n   - Alibaba Coding Plan is direct Alibaba Model Studio\u002FCoding Plan, not Command Code, OpenCode, 9router, or any local proxy.\n   - Verify against the authoritative source or current configured provider before stating a model is available.\n\n3. **Delivery method**\n   - Telegram delivery means use Telegram tools\u002Fattachments.\n   - Do not invoke Resend\u002Femail rules for Telegram delivery.\n   - If the user asks for an MD file over Telegram, send the actual `.md` file attachment, not a summary.\n\n4. **Known bad recommendations**\n   - Do not recommend Meilisearch on this VPS. It has caused memory leaks \u002F RAM exhaustion \u002F daily crashes.\n   - Do not recommend local GPU model services; VPS has no GPU.\n   - Do not recommend more OCR stacks unless the user explicitly asks; OCR is already strong enough.\n   - **Command Code via Hermes**: Do not claim Kimi\u002Fcommand-code lacks Telegram, VPS access, or Hermes skills. Delivery = Hermes Gateway. VPS shell = CLI proxy :9877. Skills = read `SKILL.md` from disk (no native `skill_view`). Load `command-code-proxy` before stating capabilities.\n\n5. **Verification before claiming done**\n   - Before saying a file was sent, verify the tool result includes an attachment success.\n   - Before saying a field\u002Ffile\u002Fservice was fixed, inspect or test it.\n   - If not verified, say exactly what was done and what is unverified.\n   - **Notary-log cal host:** Before saying Settings shows tokens, \"Create my account\", or Cal UI exists, verify **deployed** bundle (`bun run build` + service restart + grep dist or live fetch). Code in repo alone is not proof Joseph sees it.\n   - Before running `verify-cal-host.mjs`, warn it **wipes** cal DB users\u002Fbookings — Joseph's phone token goes stale.\n   - Cal \"100% complete\": split Plan A MVP (paste-link + webhooks on cal host) vs OAuth vs Worker ship vs real Cal booking on user's phone.\n\n5a. **Verify before claiming ABSENCE (hard-won, recurring failure)**\n   - Do NOT say \"there is no native import \u002F endpoint \u002F feature \u002F file \u002F capability\" based on a shallow first-pass grep or memory of \"how these tools usually work.\" This pattern burned the user badly on 2026-07-25: the assistant twice made confident negative claims (\"no native 9router→OmniRoute import exists\", \"Qwen session cookies are not persisted \u002F can't be extracted\") that were both **wrong**, and the user correctly called it lying.\n   - Before claiming absence of a feature\u002Fendpoint\u002Ffile\u002Fcapability in a tool on this VPS (OmniRoute, 9router, qwen-gate, etc.):\n     1. Grep the source tree for the symbol\u002Fendpoint name (case-insensitive, both camelCase and snake_case).\n     2. Check the actual routes\u002Fendpoints the running server exposes (`curl localhost:\u003Cport>\u002Fapi\u002F...`, read the route files).\n     3. Check the actual on-disk state (SQLite tables, browser profiles, `.env`, config files) — don't answer from memory.\n     4. Only THEN say \"no, it doesn't exist\" — and cite what you checked.\n   - If you didn't do steps 1-3, say \"I need to verify\" and do them. A wrong \"no\" wastes more time than a verification pass. The user has **zero tolerance** for confident-but-unverified negative claims — repeated instances get treated as lying, not as honest mistakes.\n\n6. **Version \u002F update claims (Joe)**\n   - Do not equate \"installed ≠ latest release\" with \"you need to update\" for Hermes, mem0, QWEN-gate, or 9router unless Joe asked for upgrades or the install is broken\u002Fmissing.\n   - Hermes at `\u002Fusr\u002Flocal\u002Flib\u002Fhermes-agent` may be a custom build on purpose.\n   - If Joe says an item does not need updating, answer NO — do not override with GitHub\u002FPyPI tags.\n   - If you already gave a wrong yes\u002Fno in the same thread, correct it in one direct sentence.\n\n## Command\n\nRun:\n\n```bash\npython3 \u002Fhome\u002Fworkspace\u002FSkills\u002Fagent-accuracy-guard\u002Fscripts\u002Fpreflight.py --task \"\u003Ctask>\"\n```\n\nIt prints a checklist to follow before acting.\n","# Agent Accuracy Guard\n\nUse this before high-risk actions.\n\n## Non-negotiable checks\n\n1. **Authorization**\n   - If the user asked for a plan first, stop after the plan.\n   - Do not run agent swarm, bulk jobs, paid APIs, email, or external sends without explicit approval.\n   - \"I guess continue\" counts as approval only if the user is clearly telling you to continue the same pending action.\n\n2. **Provider\u002Fmodel verification**\n   - Never infer model availability from a local proxy unless the user explicitly said to use that proxy.\n   - Alibaba Coding Plan is direct Alibaba Model Studio\u002FCoding Plan, not Command Code, OpenCode, 9router, or any local proxy.\n   - Verify against the authoritative source or current configured provider before stating a model is available.\n\n3. **Delivery method**\n   - Telegram delivery means use Telegram tools\u002Fattachments.\n   - Do not invoke Resend\u002Femail rules for Telegram delivery.\n   - If the user asks for an MD file over Telegram, send the actual `.md` file attachment, not a summary.\n\n4. **Known bad recommendations**\n   - Do not recommend Meilisearch on this VPS. It has caused memory leaks \u002F RAM exhaustion \u002F daily crashes.\n   - Do not recommend local GPU model services; VPS has no GPU.\n   - Do not recommend more OCR stacks unless the user explicitly asks; OCR is already strong enough.\n   - **Command Code via Hermes**: Do not claim Kimi\u002Fcommand-code lacks Telegram, VPS access, or Hermes skills. Delivery = Hermes Gateway. VPS shell = CLI proxy :9877. Skills = read `SKILL.md` from disk (no native `skill_view`). Load `command-code-proxy` before stating capabilities.\n\n5. **Verification before claiming done**\n   - Before saying a file was sent, verify the tool result includes an attachment success.\n   - Before saying a field\u002Ffile\u002Fservice was fixed, inspect or test it.\n   - If not verified, say exactly what was done and what is unverified.\n   - **Notary-log cal host:** Before saying Settings shows tokens, \"Create my account\", or Cal UI exists, verify **deployed** bundle (`bun run build` + service restart + grep dist or live fetch). Code in repo alone is not proof Joseph sees it.\n   - Before running `verify-cal-host.mjs`, warn it **wipes** cal DB users\u002Fbookings — Joseph's phone token goes stale.\n   - Cal \"100% complete\": split Plan A MVP (paste-link + webhooks on cal host) vs OAuth vs Worker ship vs real Cal booking on user's phone.\n\n5a. **Verify before claiming ABSENCE (hard-won, recurring failure)**\n   - Do NOT say \"there is no native import \u002F endpoint \u002F feature \u002F file \u002F capability\" based on a shallow first-pass grep or memory of \"how these tools usually work.\" This pattern burned the user badly on 2026-07-25: the assistant twice made confident negative claims (\"no native 9router→OmniRoute import exists\", \"Qwen session cookies are not persisted \u002F can't be extracted\") that were both **wrong**, and the user correctly called it lying.\n   - Before claiming absence of a feature\u002Fendpoint\u002Ffile\u002Fcapability in a tool on this VPS (OmniRoute, 9router, qwen-gate, etc.):\n     1. Grep the source tree for the symbol\u002Fendpoint name (case-insensitive, both camelCase and snake_case).\n     2. Check the actual routes\u002Fendpoints the running server exposes (`curl localhost:\u003Cport>\u002Fapi\u002F...`, read the route files).\n     3. Check the actual on-disk state (SQLite tables, browser profiles, `.env`, config files) — don't answer from memory.\n     4. Only THEN say \"no, it doesn't exist\" — and cite what you checked.\n   - If you didn't do steps 1-3, say \"I need to verify\" and do them. A wrong \"no\" wastes more time than a verification pass. The user has **zero tolerance** for confident-but-unverified negative claims — repeated instances get treated as lying, not as honest mistakes.\n\n6. **Version \u002F update claims (Joe)**\n   - Do not equate \"installed ≠ latest release\" with \"you need to update\" for Hermes, mem0, QWEN-gate, or 9router unless Joe asked for upgrades or the install is broken\u002Fmissing.\n   - Hermes at `\u002Fusr\u002Flocal\u002Flib\u002Fhermes-agent` may be a custom build on purpose.\n   - If Joe says an item does not need updating, answer NO — do not override with GitHub\u002FPyPI tags.\n   - If you already gave a wrong yes\u002Fno in the same thread, correct it in one direct sentence.\n\n## Command\n\nRun:\n\n```bash\npython3 \u002Fhome\u002Fworkspace\u002FSkills\u002Fagent-accuracy-guard\u002Fscripts\u002Fpreflight.py --task \"\u003Ctask>\"\n```\n\nIt prints a checklist to follow before acting.\n",{},"main","3641543f81ae80504027d6b012e0dadce4e2de40","skills\u002Fagent-accuracy-guard","https:\u002F\u002Fgithub.com\u002FSillyHippy\u002Fzo-skills","https:\u002F\u002Fgithub.com\u002FSillyHippy","2026-09-18 11:42:22",{"source":7,"skillName":5,"commands":29,"prompt":39,"zipAvailable":40},[30,33,36],{"tool":31,"command":32},"npx","npx skills add SillyHippy\u002Fzo-skills --skill agent-accuracy-guard",{"tool":34,"command":35},"bunx","bunx skills add SillyHippy\u002Fzo-skills --skill agent-accuracy-guard",{"tool":37,"command":38},"pnpm","pnpm dlx skills add SillyHippy\u002Fzo-skills --skill agent-accuracy-guard","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FSillyHippy\u002Fzo-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-accuracy-guard\u002FSKILL.md",true,[42,48],{"path":43,"resourceType":44,"resourceTypeDesc":45,"sizeBytes":46,"sourceHtmlUrl":47},"CHECKLIST.md","REFERENCE","参考文档",1080,"https:\u002F\u002Fgithub.com\u002FSillyHippy\u002Fzo-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-accuracy-guard\u002FCHECKLIST.md",{"path":49,"resourceType":50,"resourceTypeDesc":51,"sizeBytes":52,"sourceHtmlUrl":53},"scripts\u002Fpreflight.py","SCRIPT","脚本",1170,"https:\u002F\u002Fgithub.com\u002FSillyHippy\u002Fzo-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-accuracy-guard\u002Fscripts\u002Fpreflight.py",[55,62,69,76,83,90],{"skillNo":56,"name":57,"nameZh":6,"description":58,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":6,"repositoryNo":11,"starsCount":12,"licenseSpdxId":13,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":17},"01M7C4RK2ZCHPKQT393RVSC6ER","zopack","Export and import zo.space route setups as shareable .zopack.md files. Use when the user wants to package their zo.space routes for sharing, or when they receive a .zopack.md file to deploy. Triggers on \"export my zo.space\", \"create a zopack\", \"import this zopack\", or any mention of .zopack.md files.","https:\u002F\u002Fgithub.com\u002FSillyHippy\u002Fzo-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzopack\u002FSKILL.md",[],[],{"skillNo":63,"name":64,"nameZh":6,"description":65,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":6,"repositoryNo":11,"starsCount":12,"licenseSpdxId":13,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":17},"01M7C4REH5EJ22X3C1SF88X6KQ","zoho-mail-inbox","Full inbox management for Zoho Mail (info@justlegalsolutions.org) — list unread, mark all as read, delete by sender\u002Fkeyword, search, send from aliases, list folders. Fully self-contained — zero setup needed.","https:\u002F\u002Fgithub.com\u002FSillyHippy\u002Fzo-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzoho-mail-inbox\u002FSKILL.md",[],[],{"skillNo":70,"name":71,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":6,"repositoryNo":11,"starsCount":12,"licenseSpdxId":13,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":17},"01M7C4RAJF1N9R8ZRT25GMHFBF","weekly-meeting-reminder","Get an email every Sunday evening listing your meetings for the coming week.","https:\u002F\u002Fgithub.com\u002FSillyHippy\u002Fzo-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzo-weekly-meeting-reminder\u002FSKILL.md",[],[],{"skillNo":77,"name":78,"nameZh":6,"description":79,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":6,"repositoryNo":11,"starsCount":12,"licenseSpdxId":13,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":17},"01M7C4R65Q9T6TPBRFQPWJSW4Z","twitter","Let Zo use your X (Twitter) account","https:\u002F\u002Fgithub.com\u002FSillyHippy\u002Fzo-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzo-twitter\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":6,"repositoryNo":11,"starsCount":12,"licenseSpdxId":13,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":17},"01M7C4R358YM59EZRM3SW763VK","zo-tools","Internal process tools for Zo and Hermes — local search, document processing, OCR, and embeddings. All services run on localhost, zero HTTP slots consumed.","https:\u002F\u002Fgithub.com\u002FSillyHippy\u002Fzo-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzo-tools\u002FSKILL.md",[],[],{"skillNo":91,"name":92,"nameZh":6,"description":93,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":6,"repositoryNo":11,"starsCount":12,"licenseSpdxId":13,"licenseFileUrl":6,"sourceHtmlUrl":94,"categories":95,"keywords":96,"publishedAt":17},"01M7C4R0J4C3RQWKZ5W5W3CZ3X","ssh-setup","Sets up secure remote access to your Zo Computer so you can connect from anywhere","https:\u002F\u002Fgithub.com\u002FSillyHippy\u002Fzo-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzo-ssh-setup\u002FSKILL.md",[],[]]