[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7F2MC8MB6H05K4PHY9P8DM9":3,"skill-same:01M7F2MC8MB6H05K4PHY9P8DM9":45},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":14,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":29,"updatedAt":30,"install":31,"resources":44},"01M7F2MC8MB6H05K4PHY9P8DM9","grillspec-grill",null,"Use when a change involves decisions the user should make - before writing a spec, plan or code for a feature, module, interface, schema or infrastructure change - and when the user says grill, interview me, let's plan, or asks how to build something. Interviews until nothing is left to guess and keeps a decision log that survives context loss.","forbxpg\u002Fgrillspec","forbxpg","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F144426975?v=4","01M7F2KNWHMXF9R8YPQNDQPECB","Kirgizov Temurmalik","01M7F2KNX4T75NN17D5C25KKKZ",0,"MIT","https:\u002F\u002Fgithub.com\u002Fforbxpg\u002Fgrillspec\u002Fblob\u002Fmain\u002Fskills\u002Fgrillspec-grill\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: grillspec-grill\ndescription: Use when a change involves decisions the user should make - before writing a spec, plan or code for a feature, module, interface, schema or infrastructure change - and when the user says grill, interview me, let's plan, or asks how to build something. Interviews until nothing is left to guess and keeps a decision log that survives context loss.\n---\n\n# Grill the user before building\n\nA design built on a guess costs more than ten extra questions. Interview the user\nuntil you could write the spec without guessing anywhere, and write every answer\nto the decision log as you go: the log, not your memory, is what the spec is\nwritten from.\n\n`grillspec` below means the command on the `grillspec CLI:` line of your\nsession-start context (or `uvx grillspec`).\n\n## 1. Start\n\n1. Read the session-start status line (or run `grillspec status --json`).\n   - Phase `grill`: you are resuming. Read the log at its `log=` path, then go\n     straight to step 3 with its Open items. Do not re-ask Settled ones, and do not\n     re-explore the code you already explored for them.\n   - Another change is active in `grill`, `spec`, `plan` or `execute` and the\n     request is not part of it: name it and its phase and ask — continue it \u002F start\n     the new one here (the old one stays; `grillspec use \u003Cid>` returns to it) \u002F\n     start the new one in a separate worktree (`git worktree add ..\u002F\u003Crepo>-\u003Cslug>`,\n     then continue there).\n2. Ask the **setup question**, one message:\n   - a slug for the change (lowercase kebab-case, proposed from the request);\n   - the depth — recommend by stakes: `quick` for a small, reversible change;\n     `standard` for a feature; `deep` for anything hard to reverse (schema, public\n     API, infrastructure, security);\n   - a council, only if a `grillspec-council` skill is installed and\n     `council.enabled` is `ask` in the config. Name the members it would use (see\n     that skill) and the cost: `≈ \u003Cdecisions> × \u003Cmembers> calls (standard ≈ 4\n     decisions, deep ≈ 15); external calls spend your vendor quotas`.\n3. Run `grillspec new \u003Cslug> --depth \u003Cdepth>` (add `--council` when chosen, and\n   `--switch` when the user chose to start it while another change is active).\n   If the depth, council or council members differ from the configured default,\n   offer once to make them the default in the user's config.\n\n## 2. Build the decision tree\n\nMap every decision the change needs, with the decisions that hang off it. The\n**frontier** is every decision whose prerequisites are settled. Write the whole\nfrontier under `### Open` before asking anything:\n\n```markdown\n- [ ] Q3 Retry policy for failed webhooks (after D2)\n```\n\n**Facts are yours, decisions are theirs.** Look facts up: files, git history,\nconfig, docs, a `--help`. A quick lookup you do inline; a broad exploration you\nhand to a subagent when the harness has one, and keep asking the questions that\ndo not depend on it. Never ask the user what you could have checked. A choice\nbetween options is a decision even when one option looks obvious — ask it.\n\n## 3. Ask\n\n| Depth | How | Budget |\n|---|---|---|\n| quick | Rounds in plain text: the whole frontier, numbered, each with `➡️` your recommendation, worded so \"yes\" accepts all of them | ~5 decisions |\n| standard | One decision at a time; up to 4 small independent ones together | ~15 decisions |\n| deep | One decision at a time | none |\n\nFor standard and deep use the harness's multiple-choice tool when it has one\n(see `using-grillspec\u002Freferences\u002F`), otherwise numbered options in text.\n\nEvery decision you put carries: your recommendation first, marked\n`(Recommended)`; the real trade-off of each option; whether it is cheap or costly\nto reverse. With a council (`council=yes` in the metadata), follow\n`grillspec-council` — on standard for root and hard-to-reverse decisions, on deep\nfor every decision.\n\n## 4. Log every answer — immediately\n\nIn the same turn as your next question, move the item from Open to Settled:\n\n```markdown\n- [x] D1 Auth — **JWT + refresh tokens** · user · alternatives: server sessions (simpler revoke) · reversible: costly\n  > \"the mobile client has no cookies\"\n```\n\n- Source, exactly one of: `user` (they chose), `accepted` (they said yes to your\n  recommendation), `delegated` (\"you decide\": take your recommendation, do not ask\n  again), `agent's call` (only for choices with no effect on the product, such as\n  an internal name), `accepted without discussion` (see Budget).\n- \"Skip\" \u002F \"later\": leave it Open and append `· deferred`.\n- Quote the user's own words when they answered in their own words.\n- Changed mind: never edit a Settled line. Add a new entry and mark the old one\n  `- [~] D1 … superseded by D6`.\n- Write in the project's artifact language; OpenSpec keywords stay English.\n- Never write secrets, tokens or passwords into the log: write `\u003Credacted>`.\n- Bump `asked=` in the `\u003C!-- grillspec: … -->` line when you put a decision to the\n  user — not when you add it to Open.\n- New branches that an answer opens go under Open right away.\n\n## 5. Budget\n\nThe budget counts decisions put to the user, batched or not; follow-up questions\nabout the same decision and fact lookups do not count. When `asked` reaches it,\nlist every remaining Open item with your recommendation and ask: continue, or\naccept these? Accepted ones are logged as `accepted without discussion`.\n\n## 6. Split\n\nIf the tree turns out to hold independent changes, name them and offer to split.\nOn yes, create each with `grillspec new \u003Cslug> --switch`, move its Open items to\nits log, return with `grillspec use \u003Cid>`, and finish one change at a time.\n\n## 7. Finish\n\nWhen Open holds nothing but explicitly deferred items:\n\n1. Play back the shared understanding in 3–10 lines. List separately every\n   `agent's call`, `delegated` and `deferred` item.\n2. Ask: \"Do we share the same understanding?\" Only an explicit yes counts.\n3. On yes, run `grillspec approve grill`.\n4. Name the log file and suggest committing it — never commit it yourself.\n5. Continue with the `grillspec-specs` skill. If it is not installed, say so and\n   stop: the log is complete and the phase is now `spec`.\n\n## Red flags\n\n| Thought | Reality |\n|---|---|\n| \"This one is obvious.\" | Obvious is where unchecked assumptions live. Ask it. |\n| \"They're in a hurry, I'll decide.\" | Name the decision in one sentence and ask. One question beats one rework. |\n| \"I'll ask that later.\" | Later is after the guess is built in. Put it under Open now. |\n| \"I'll write the log at the end.\" | The end comes after compaction. Write it now. |\n| \"That's a fact, not a decision.\" | If two reasonable people could choose differently, it is a decision. |\n| \"They said yes earlier, so this is approved too.\" | Each decision and the final confirmation need their own yes. |\n","# Grill the user before building\n\nA design built on a guess costs more than ten extra questions. Interview the user\nuntil you could write the spec without guessing anywhere, and write every answer\nto the decision log as you go: the log, not your memory, is what the spec is\nwritten from.\n\n`grillspec` below means the command on the `grillspec CLI:` line of your\nsession-start context (or `uvx grillspec`).\n\n## 1. Start\n\n1. Read the session-start status line (or run `grillspec status --json`).\n   - Phase `grill`: you are resuming. Read the log at its `log=` path, then go\n     straight to step 3 with its Open items. Do not re-ask Settled ones, and do not\n     re-explore the code you already explored for them.\n   - Another change is active in `grill`, `spec`, `plan` or `execute` and the\n     request is not part of it: name it and its phase and ask — continue it \u002F start\n     the new one here (the old one stays; `grillspec use \u003Cid>` returns to it) \u002F\n     start the new one in a separate worktree (`git worktree add ..\u002F\u003Crepo>-\u003Cslug>`,\n     then continue there).\n2. Ask the **setup question**, one message:\n   - a slug for the change (lowercase kebab-case, proposed from the request);\n   - the depth — recommend by stakes: `quick` for a small, reversible change;\n     `standard` for a feature; `deep` for anything hard to reverse (schema, public\n     API, infrastructure, security);\n   - a council, only if a `grillspec-council` skill is installed and\n     `council.enabled` is `ask` in the config. Name the members it would use (see\n     that skill) and the cost: `≈ \u003Cdecisions> × \u003Cmembers> calls (standard ≈ 4\n     decisions, deep ≈ 15); external calls spend your vendor quotas`.\n3. Run `grillspec new \u003Cslug> --depth \u003Cdepth>` (add `--council` when chosen, and\n   `--switch` when the user chose to start it while another change is active).\n   If the depth, council or council members differ from the configured default,\n   offer once to make them the default in the user's config.\n\n## 2. Build the decision tree\n\nMap every decision the change needs, with the decisions that hang off it. The\n**frontier** is every decision whose prerequisites are settled. Write the whole\nfrontier under `### Open` before asking anything:\n\n```markdown\n- [ ] Q3 Retry policy for failed webhooks (after D2)\n```\n\n**Facts are yours, decisions are theirs.** Look facts up: files, git history,\nconfig, docs, a `--help`. A quick lookup you do inline; a broad exploration you\nhand to a subagent when the harness has one, and keep asking the questions that\ndo not depend on it. Never ask the user what you could have checked. A choice\nbetween options is a decision even when one option looks obvious — ask it.\n\n## 3. Ask\n\n| Depth | How | Budget |\n|---|---|---|\n| quick | Rounds in plain text: the whole frontier, numbered, each with `➡️` your recommendation, worded so \"yes\" accepts all of them | ~5 decisions |\n| standard | One decision at a time; up to 4 small independent ones together | ~15 decisions |\n| deep | One decision at a time | none |\n\nFor standard and deep use the harness's multiple-choice tool when it has one\n(see `using-grillspec\u002Freferences\u002F`), otherwise numbered options in text.\n\nEvery decision you put carries: your recommendation first, marked\n`(Recommended)`; the real trade-off of each option; whether it is cheap or costly\nto reverse. With a council (`council=yes` in the metadata), follow\n`grillspec-council` — on standard for root and hard-to-reverse decisions, on deep\nfor every decision.\n\n## 4. Log every answer — immediately\n\nIn the same turn as your next question, move the item from Open to Settled:\n\n```markdown\n- [x] D1 Auth — **JWT + refresh tokens** · user · alternatives: server sessions (simpler revoke) · reversible: costly\n  > \"the mobile client has no cookies\"\n```\n\n- Source, exactly one of: `user` (they chose), `accepted` (they said yes to your\n  recommendation), `delegated` (\"you decide\": take your recommendation, do not ask\n  again), `agent's call` (only for choices with no effect on the product, such as\n  an internal name), `accepted without discussion` (see Budget).\n- \"Skip\" \u002F \"later\": leave it Open and append `· deferred`.\n- Quote the user's own words when they answered in their own words.\n- Changed mind: never edit a Settled line. Add a new entry and mark the old one\n  `- [~] D1 … superseded by D6`.\n- Write in the project's artifact language; OpenSpec keywords stay English.\n- Never write secrets, tokens or passwords into the log: write `\u003Credacted>`.\n- Bump `asked=` in the `\u003C!-- grillspec: … -->` line when you put a decision to the\n  user — not when you add it to Open.\n- New branches that an answer opens go under Open right away.\n\n## 5. Budget\n\nThe budget counts decisions put to the user, batched or not; follow-up questions\nabout the same decision and fact lookups do not count. When `asked` reaches it,\nlist every remaining Open item with your recommendation and ask: continue, or\naccept these? Accepted ones are logged as `accepted without discussion`.\n\n## 6. Split\n\nIf the tree turns out to hold independent changes, name them and offer to split.\nOn yes, create each with `grillspec new \u003Cslug> --switch`, move its Open items to\nits log, return with `grillspec use \u003Cid>`, and finish one change at a time.\n\n## 7. Finish\n\nWhen Open holds nothing but explicitly deferred items:\n\n1. Play back the shared understanding in 3–10 lines. List separately every\n   `agent's call`, `delegated` and `deferred` item.\n2. Ask: \"Do we share the same understanding?\" Only an explicit yes counts.\n3. On yes, run `grillspec approve grill`.\n4. Name the log file and suggest committing it — never commit it yourself.\n5. Continue with the `grillspec-specs` skill. If it is not installed, say so and\n   stop: the log is complete and the phase is now `spec`.\n\n## Red flags\n\n| Thought | Reality |\n|---|---|\n| \"This one is obvious.\" | Obvious is where unchecked assumptions live. Ask it. |\n| \"They're in a hurry, I'll decide.\" | Name the decision in one sentence and ask. One question beats one rework. |\n| \"I'll ask that later.\" | Later is after the guess is built in. Put it under Open now. |\n| \"I'll write the log at the end.\" | The end comes after compaction. Write it now. |\n| \"That's a fact, not a decision.\" | If two reasonable people could choose differently, it is a decision. |\n| \"They said yes earlier, so this is approved too.\" | Each decision and the final confirmation need their own yes. |\n",{"name":5,"description":7},"main","348b0fb01ff821b20f8a5c13878439f18ca9ee5e","skills\u002Fgrillspec-grill","https:\u002F\u002Fgithub.com\u002Fforbxpg\u002Fgrillspec","https:\u002F\u002Fgithub.com\u002Fforbxpg","Pretty chill irl but incredibly turbulent in the coding steam","2026-10-21 18:43:44",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add forbxpg\u002Fgrillspec --skill grillspec-grill",{"tool":37,"command":38},"bunx","bunx skills add forbxpg\u002Fgrillspec --skill grillspec-grill",{"tool":40,"command":41},"pnpm","pnpm dlx skills add forbxpg\u002Fgrillspec --skill grillspec-grill","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fforbxpg\u002Fgrillspec\u002Fblob\u002Fmain\u002Fskills\u002Fgrillspec-grill\u002FSKILL.md",true,[],[46,52,57,64,71,78],{"skillNo":47,"name":48,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":49,"categories":50,"keywords":51,"publishedAt":19},"01M7F2PAK6YPZZ3J2DHYTMSAE9","graders","https:\u002F\u002Fgithub.com\u002Fforbxpg\u002Fgrillspec\u002Fblob\u002Fmain\u002Ftests\u002Fevals\u002Fgrill-triggers\u002Fgraders\u002Fskill.md",[],[],{"skillNo":53,"name":48,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":54,"categories":55,"keywords":56,"publishedAt":19},"01M7F2P8RZ76WHAB41AN13JPZT","https:\u002F\u002Fgithub.com\u002Fforbxpg\u002Fgrillspec\u002Fblob\u002Fmain\u002Ftests\u002Fevals\u002Farchive-asks\u002Fgraders\u002Fskill.md",[],[],{"skillNo":58,"name":59,"nameZh":6,"description":60,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":19},"01M7F2P459S5T73RNMGXNPSTAT","using-grillspec","Use when starting any conversation, and before acting on any request that involves choosing between options - routes work through grilling, specs, plans and execution so the agent never guesses what the user would decide.","https:\u002F\u002Fgithub.com\u002Fforbxpg\u002Fgrillspec\u002Fblob\u002Fmain\u002Fskills\u002Fusing-grillspec\u002FSKILL.md",[],[],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":19},"01M7F2NZJEKXP7B1YZKVWDFJSA","grillspec-worktrees","Use when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback","https:\u002F\u002Fgithub.com\u002Fforbxpg\u002Fgrillspec\u002Fblob\u002Fmain\u002Fskills\u002Fgrillspec-worktrees\u002FSKILL.md",[],[],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":19},"01M7F2NTZFV9EVXEH7VEFWM58W","grillspec-verification","Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always","https:\u002F\u002Fgithub.com\u002Fforbxpg\u002Fgrillspec\u002Fblob\u002Fmain\u002Fskills\u002Fgrillspec-verification\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":19},"01M7F2NQ2YHVGVTJZJSRFH8GQQ","grillspec-tdd","Use when implementing any feature or bugfix, before writing implementation code","https:\u002F\u002Fgithub.com\u002Fforbxpg\u002Fgrillspec\u002Fblob\u002Fmain\u002Fskills\u002Fgrillspec-tdd\u002FSKILL.md",[],[]]