[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CPADA4J371APS3SESABC7M":3,"skill-same:01M7CPADA4J371APS3SESABC7M":49},{"skillNo":4,"name":5,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":28,"commitSha":29,"skillDirectory":30,"forksCount":13,"repositoryHtmlUrl":31,"repositoryHomepageUrl":6,"ownerHtmlUrl":32,"ownerBlogUrl":33,"ownerDescription":34,"updatedAt":35,"install":36,"resources":48},"01M7CPADA4J371APS3SESABC7M","refine",null,"markosAMO\u002FAlfred","markosAMO","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F51459843?v=4","01M7CP9J22V3QPXPRGE3MJZ091","Markos Moscoso Ocampo","01M7CP9J33G8P7T5TBN51PST9G",0,"MIT","https:\u002F\u002Fgithub.com\u002FmarkosAMO\u002FAlfred\u002Fblob\u002Fmain\u002Fskills\u002Frefine\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: refine\nmode: interactive\nskippable: true\nnext: [research, spec]\n---\n\n# refine\n\nInterview the user until the request is understood well enough to specify. Write nothing\nelse: no specification, no design, no code.\n\nThis phase exists because a request arrives shorter than the thing being requested. \"Add\nGoogle login\" is eight words describing a week of work, and every gap left in it is filled\nby the model guessing.\n\n## When the orchestrator routes here\n\nA request needs this phase when any of these hold. Otherwise the route can start past it.\n\n- The failure behaviour is unstated. The request says what should happen when it works.\n- A domain term appears that is not defined in `architecture.md` or memory.\n- More than one component or repository is affected.\n- It requires an architectural decision that has not been made.\n- There is no way to tell, from the request alone, whether the result is correct.\n\nA request carrying none of these is already specifiable, and interviewing the user about\nit wastes their time.\n\nWhen the request arrives as a tracker card, the same criteria apply to the card body.\n\nThe decision is proposed and confirmed, never applied silently, and the orchestrator names\nthe signals it matched and lets the user take the shorter answer. Those mechanics hold under\nevery workflow and are in `skills\u002F_shared\u002Frouting.md`.\n\nWhat matching them is worth, and where the shorter answer goes instead, is the running\nworkflow's and is written in its rules file. The list above is what this phase is for; the\nroute it argues for belongs to the workflow that declares the routes.\n\n## Materialise the inputs first\n\nBefore anything else, bring in whatever arrived as a reference: tracker cards, URLs,\ndocuments in other systems.\n\n```\nread each source once\nwrite it in full to .alfred\u002Fchanges\u002F{change}\u002Finputs\u002F\nindex it under alfred\u002F{change}\u002Finput\u002F{slug}\n```\n\nFollow `skills\u002F_shared\u002Fexternal-inputs.md`. Nothing after this phase fetches anything: the\ninterview, `spec`, `design` and every subagent read the materialised text.\n\n## Before asking anything\n\nRead what is already known, and never ask for it.\n\n```\ninputs\u002F                     the material just materialised\narchitecture, conventions   when the workflow hands them\nrecall                      the types the workflow declares, for prior decisions and related changes\n```\n\nAsking which database the project uses, when `architecture.md` states it, teaches the user\nthat answering carefully is pointless.\n\nOpen every round by naming what was assumed from these sources, so a wrong assumption is\ncorrected instead of inherited.\n\n## The interview\n\nQuestions are asked in rounds, grouped by topic. One `ask()` per round carrying three or\nfour questions, never one question at a time.\n\nEach round is answered before the next is asked. Later rounds depend on earlier answers:\nasking about integration before the scope is settled produces questions about work that\nturns out to be excluded.\n\n### Round 1 — Problem and scope\n\n- What problem does this solve, and for whom?\n- What does the user do today instead?\n- What is explicitly out of scope?\n- Is this new behaviour, or a change to behaviour that already exists?\n\n### Round 2 — Behaviour\n\n- What happens when it works? Walk through it.\n- What can go wrong, and what should happen then?\n- Which edge cases matter, and which are acceptable to ignore?\n- What must never happen?\n\nThe failure path is not optional. A round that produces only the happy path is not\nfinished, because `testing-protocol.md` requires both and the scenarios come from here.\n\n### Round 3 — Integration and direction\n\n- Which parts of the system does this touch?\n- Which external services are involved?\n- What data is read, written or stored?\n- Is there an approach you already want, or one you want ruled out?\n\nThe last question is the only one that collects a technical decision, and it collects the\nuser's, never the model's. An answer to it is a constraint on `design`, not a design.\n\nLeave it unanswered rather than filling it in. A direction the user did not state is a\ndecision `design` should be making with the specification in front of it.\n\nFor multi-repository work, this round determines which repositories are affected and is\nthe input to workspace `design`.\n\n### Round 4 — Constraints and acceptance\n\n- How will we know it is done?\n- Are there performance, security or compliance constraints?\n- Is there a deadline or a dependency on other work?\n- What would make you reject this once built?\n\n## Stopping\n\nStop when a round produces nothing that changes the proposal. Continuing past that point\ncollects detail that belongs in `design`.\n\nStop early when the user says to. A partial proposal with its gaps marked is more useful\nthan a complete one built on invented answers.\n\nNever fill a gap by assuming. An unanswered question is written into the proposal as an\nopen question and carried into `spec`.\n\n## What it writes\n\n```markdown\n## Problem\n## Proposed change\n## Scope\n### Included\n### Excluded\n## Behaviour\n### Happy path\n### Failure paths\n### Edge cases\n## Integration\n### Components affected\n### External services\n### Data\n## Architectural direction\n## Constraints\n## Acceptance\n## Open questions\n## Assumptions\n```\n\n`Architectural direction` holds only what the user decided, each line with the reason they\ngave. It is empty when they decided nothing, and an empty section is the normal case.\n\n```markdown\n## Architectural direction\n- Events, not polling. The consumer cannot be blocked by the producer.\n- Reuse the existing notification service rather than adding a second one.\n```\n\nThese are constraints `design` inherits, not a design. They name a direction and a reason;\nthey name no interface, no schema and no library.\n\n`Assumptions` records what was taken from `architecture.md` or memory rather than asked.\n`spec` treats those as inherited, not as decided, and a wrong one surfaces there instead of\nin the code.\n\n## Completion\n\nFollow `skills\u002F_shared\u002Fphase-protocol.md`: write the artifact with its `## Handoff`, update state, notify `phase_completed`.\n\nThe proposal is prose describing intent. It contains no `SHALL`, no scenarios and no delta\nsections. Those belong to `spec`, and writing them here produces a requirement nobody\nreviewed as one.\n","# refine\n\nInterview the user until the request is understood well enough to specify. Write nothing\nelse: no specification, no design, no code.\n\nThis phase exists because a request arrives shorter than the thing being requested. \"Add\nGoogle login\" is eight words describing a week of work, and every gap left in it is filled\nby the model guessing.\n\n## When the orchestrator routes here\n\nA request needs this phase when any of these hold. Otherwise the route can start past it.\n\n- The failure behaviour is unstated. The request says what should happen when it works.\n- A domain term appears that is not defined in `architecture.md` or memory.\n- More than one component or repository is affected.\n- It requires an architectural decision that has not been made.\n- There is no way to tell, from the request alone, whether the result is correct.\n\nA request carrying none of these is already specifiable, and interviewing the user about\nit wastes their time.\n\nWhen the request arrives as a tracker card, the same criteria apply to the card body.\n\nThe decision is proposed and confirmed, never applied silently, and the orchestrator names\nthe signals it matched and lets the user take the shorter answer. Those mechanics hold under\nevery workflow and are in `skills\u002F_shared\u002Frouting.md`.\n\nWhat matching them is worth, and where the shorter answer goes instead, is the running\nworkflow's and is written in its rules file. The list above is what this phase is for; the\nroute it argues for belongs to the workflow that declares the routes.\n\n## Materialise the inputs first\n\nBefore anything else, bring in whatever arrived as a reference: tracker cards, URLs,\ndocuments in other systems.\n\n```\nread each source once\nwrite it in full to .alfred\u002Fchanges\u002F{change}\u002Finputs\u002F\nindex it under alfred\u002F{change}\u002Finput\u002F{slug}\n```\n\nFollow `skills\u002F_shared\u002Fexternal-inputs.md`. Nothing after this phase fetches anything: the\ninterview, `spec`, `design` and every subagent read the materialised text.\n\n## Before asking anything\n\nRead what is already known, and never ask for it.\n\n```\ninputs\u002F                     the material just materialised\narchitecture, conventions   when the workflow hands them\nrecall                      the types the workflow declares, for prior decisions and related changes\n```\n\nAsking which database the project uses, when `architecture.md` states it, teaches the user\nthat answering carefully is pointless.\n\nOpen every round by naming what was assumed from these sources, so a wrong assumption is\ncorrected instead of inherited.\n\n## The interview\n\nQuestions are asked in rounds, grouped by topic. One `ask()` per round carrying three or\nfour questions, never one question at a time.\n\nEach round is answered before the next is asked. Later rounds depend on earlier answers:\nasking about integration before the scope is settled produces questions about work that\nturns out to be excluded.\n\n### Round 1 — Problem and scope\n\n- What problem does this solve, and for whom?\n- What does the user do today instead?\n- What is explicitly out of scope?\n- Is this new behaviour, or a change to behaviour that already exists?\n\n### Round 2 — Behaviour\n\n- What happens when it works? Walk through it.\n- What can go wrong, and what should happen then?\n- Which edge cases matter, and which are acceptable to ignore?\n- What must never happen?\n\nThe failure path is not optional. A round that produces only the happy path is not\nfinished, because `testing-protocol.md` requires both and the scenarios come from here.\n\n### Round 3 — Integration and direction\n\n- Which parts of the system does this touch?\n- Which external services are involved?\n- What data is read, written or stored?\n- Is there an approach you already want, or one you want ruled out?\n\nThe last question is the only one that collects a technical decision, and it collects the\nuser's, never the model's. An answer to it is a constraint on `design`, not a design.\n\nLeave it unanswered rather than filling it in. A direction the user did not state is a\ndecision `design` should be making with the specification in front of it.\n\nFor multi-repository work, this round determines which repositories are affected and is\nthe input to workspace `design`.\n\n### Round 4 — Constraints and acceptance\n\n- How will we know it is done?\n- Are there performance, security or compliance constraints?\n- Is there a deadline or a dependency on other work?\n- What would make you reject this once built?\n\n## Stopping\n\nStop when a round produces nothing that changes the proposal. Continuing past that point\ncollects detail that belongs in `design`.\n\nStop early when the user says to. A partial proposal with its gaps marked is more useful\nthan a complete one built on invented answers.\n\nNever fill a gap by assuming. An unanswered question is written into the proposal as an\nopen question and carried into `spec`.\n\n## What it writes\n\n```markdown\n## Problem\n## Proposed change\n## Scope\n### Included\n### Excluded\n## Behaviour\n### Happy path\n### Failure paths\n### Edge cases\n## Integration\n### Components affected\n### External services\n### Data\n## Architectural direction\n## Constraints\n## Acceptance\n## Open questions\n## Assumptions\n```\n\n`Architectural direction` holds only what the user decided, each line with the reason they\ngave. It is empty when they decided nothing, and an empty section is the normal case.\n\n```markdown\n## Architectural direction\n- Events, not polling. The consumer cannot be blocked by the producer.\n- Reuse the existing notification service rather than adding a second one.\n```\n\nThese are constraints `design` inherits, not a design. They name a direction and a reason;\nthey name no interface, no schema and no library.\n\n`Assumptions` records what was taken from `architecture.md` or memory rather than asked.\n`spec` treats those as inherited, not as decided, and a wrong one surfaces there instead of\nin the code.\n\n## Completion\n\nFollow `skills\u002F_shared\u002Fphase-protocol.md`: write the artifact with its `## Handoff`, update state, notify `phase_completed`.\n\nThe proposal is prose describing intent. It contains no `SHALL`, no scenarios and no delta\nsections. Those belong to `spec`, and writing them here produces a requirement nobody\nreviewed as one.\n",{"mode":23,"name":5,"next":24,"skippable":27},"interactive",[25,26],"research","spec",true,"main","e68ffe68ac604c91a9738f0ce2acc1c7585f88f2","skills\u002Frefine","https:\u002F\u002Fgithub.com\u002FmarkosAMO\u002FAlfred","https:\u002F\u002Fgithub.com\u002FmarkosAMO","marmoscoso1@gmail.com","Dev","2026-10-13 14:30:03",{"source":7,"skillName":5,"commands":37,"prompt":47,"zipAvailable":27},[38,41,44],{"tool":39,"command":40},"npx","npx skills add markosAMO\u002FAlfred --skill refine",{"tool":42,"command":43},"bunx","bunx skills add markosAMO\u002FAlfred --skill refine",{"tool":45,"command":46},"pnpm","pnpm dlx skills add markosAMO\u002FAlfred --skill refine","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FmarkosAMO\u002FAlfred\u002Fblob\u002Fmain\u002Fskills\u002Frefine\u002FSKILL.md",[],[50,56,62,68,73,79],{"skillNo":51,"name":52,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":53,"categories":54,"keywords":55,"publishedAt":18},"01M7CPAPB4796E8W31MQFMTC9N","{phase=null}","https:\u002F\u002Fgithub.com\u002FmarkosAMO\u002FAlfred\u002Fblob\u002Fmain\u002Ftemplates\u002Fworkflow\u002Fskills\u002Fexample\u002FSKILL.md",[],[],{"skillNo":57,"name":58,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":18},"01M7CPAMS85JH7H8X17G9PM3YB","verify","https:\u002F\u002Fgithub.com\u002FmarkosAMO\u002FAlfred\u002Fblob\u002Fmain\u002Fskills\u002Fverify\u002FSKILL.md",[],[],{"skillNo":63,"name":64,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":65,"categories":66,"keywords":67,"publishedAt":18},"01M7CPAK5CR387ZSMVAH47CRW4","tasks","https:\u002F\u002Fgithub.com\u002FmarkosAMO\u002FAlfred\u002Fblob\u002Fmain\u002Fskills\u002Ftasks\u002FSKILL.md",[],[],{"skillNo":69,"name":26,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":18},"01M7CPAHZ9ZD1DNM3S5PS2RS9E","https:\u002F\u002Fgithub.com\u002FmarkosAMO\u002FAlfred\u002Fblob\u002Fmain\u002Fskills\u002Fspec\u002FSKILL.md",[],[],{"skillNo":74,"name":75,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":18},"01M7CPAGF0MDNQKSCE61F5QJ6C","review","https:\u002F\u002Fgithub.com\u002FmarkosAMO\u002FAlfred\u002Fblob\u002Fmain\u002Fskills\u002Freview\u002FSKILL.md",[],[],{"skillNo":80,"name":25,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":18},"01M7CPAF9ETJMQVXT4Q4TJQ2HN","https:\u002F\u002Fgithub.com\u002FmarkosAMO\u002FAlfred\u002Fblob\u002Fmain\u002Fskills\u002Fresearch\u002FSKILL.md",[],[]]