[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7A7R8JAX3174FMT08DWS473":3,"skill-same:01M7A7R8JAX3174FMT08DWS473":60},{"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":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":30,"ownerDescription":6,"updatedAt":31,"install":32,"resources":45},"01M7A7R8JAX3174FMT08DWS473","groom",null,"Grooms a tracker row toward Ready by forcing the judgement a Ready linter cannot make — every pointer re-read against the tree, every number reproduced, the mechanism decided, the tests shown to discriminate — then validates it with an isolated, report-only cold read before promotion. Use whenever asked to groom, refine, ready, promote or pressure-test a row or a milestone (\"groom M2 to ready\", \"get CLOUD-N ready\", \"is this row ready?\"), and before moving any row into Todo.","button-inc\u002Fbatten","button-inc","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F62446232?v=4","01M7A7QTRSW2RGC94TEV2DFHXQ","Button Inc.","01M7A7QTT73DN15FTTW2VX074X",1,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fbutton-inc\u002Fbatten\u002Fblob\u002Fmain\u002Fskills\u002Fgroom\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: groom\ndescription: Grooms a tracker row toward Ready by forcing the judgement a Ready linter cannot make — every pointer re-read against the tree, every number reproduced, the mechanism decided, the tests shown to discriminate — then validates it with an isolated, report-only cold read before promotion. Use whenever asked to groom, refine, ready, promote or pressure-test a row or a milestone (\"groom M2 to ready\", \"get CLOUD-N ready\", \"is this row ready?\"), and before moving any row into Todo.\n---\n\n# Grooming a row to Ready\n\nA Ready linter answers \"is the block well-formed\". It never answers \"is the row\ntrue\". Every defect this repository has measured in a groomed row was\n**well-formed and false**: a symbol that does not exist, a store census never\ntaken, a count with no command, a mechanism deferred behind a gate name, a test\nfile picked off a glob, a blocker that was never a blocker. So the linter's\nexit 0 is where grooming starts, never where it ends.\n\nThe deliverable of a grooming session is the **board**. A grooming session\nnever claims, readies or lands (CLOUD-1568).\n\n## The ladder — run it per row, cheapest rung first\n\nCopy this checklist and tick it per row. A row that fails a rung is fixed\nbefore the next, costlier rung is spent on it.\n\n```\nRow CLOUD-N:\n- [ ] L0 floor: `batten ready lint --issue CLOUD-N` findings fixed\n- [ ] L1 claims: every pointer, number and board claim resolved (below)\n- [ ] L1 decision: mechanism decided, tests discriminate, prose agrees with §7\n- [ ] L2 cold read: isolated validator report, every finding fixed or answered\n- [ ] promote, or leave in Backlog with the open question written on the row\n```\n\n### L0 — the floor\n\nRead the row through the verb (`batten mcp call Linear get_issue '{\"id\":\"CLOUD-N\"}'`),\nthen run `batten ready lint --issue CLOUD-N` and fix what it reports. Its exit 0\nmeans \"no known shape defect\". Never report it as \"ready\".\n\n### L1 — resolve every claim, in this session\n\nList every claim the row makes, then resolve each with the instrument\n`rules\u002Fscanning.md` names for its class. Do not reason from memory; a claim you\ndid not look up is unresolved.\n\n1. **Pointers.** Every path, `path::symbol`, rule id, task name, `input.*` key,\n   test name. Does it exist at `origin\u002Fmain`, and does it do what the row says?\n   A symbol that exists but does something else is a false premise.\n2. **Contents and live behaviour.** A claim about what a store, a session or a\n   queue CONTAINS is answered by counting it, not by reading the writer\n   (CLOUD-1523: correct in code, zero live instances).\n3. **Numbers.** Every figure names the command or fixture that reproduces it.\n   Re-run it. A number you cannot reproduce is removed or re-measured\n   (CLOUD-1522's 19,348 was a double count).\n4. **Board claims.** Blockers, \"decided on\", \"duplicate of\", a status. Search\n   the board before trusting any of them (`rules\u002Fscanning.md` row four). Look\n   for a refutation that exists elsewhere and was never written to this row.\n\nThen the decision checks, which are about the row against itself:\n\n5. **Decided, not deferred.** \"Needs a mechanism\", \"to be designed\", \"a gate\n   will decide\" beside a well-formed block is a deferral wearing a status\n   (CLOUD-1567). If the evidence in hand decides it, decide it in the row. If it\n   genuinely does not, write the open question and leave the row in Backlog.\n6. **Tests that discriminate.** Each `tests[]` entry names a real or clearly\n   new file, a case that fails without the change and passes with it, and a\n   `MUTANT \u003Cslug>|` row that `batten mutate` can reach or an honest statement\n   that it cannot (CLOUD-1526 clause 8). A file picked because it exists is the\n   defect `rules\u002Fscanning.md` measured, not a test obligation.\n7. **Prescription vs acceptance.** Does doing exactly what the row prescribes\n   satisfy the row's own §2 and §7? Where they disagree, the acceptance is the\n   spec (`mem:workflow\u002Fboard-states`).\n\nFix everything you can evidence, directly in the row: Backlog and Todo bodies\nare specs, and correcting them in place is what grooming is.\n\n### L2 — the isolated cold read\n\nThe implementer will start from nothing but the row, AGENTS.md and the repo.\nReproduce that: spawn one read-only `Explore` subagent per row, with the brief\nin `skills\u002Fgroom\u002Freferences\u002Fvalidator-brief.md` and the row body written to a\nscratch file it can Read. Rows are independent, so validators for several rows\nrun in parallel.\n\n- The validator is **report-only**. It never writes a file or the tracker. The\n  grooming session applies every fix itself, one row at a time.\n- Its value is that it did not write the row. Measured on CLOUD-1968: a fresh\n  isolated read caught every seeded defect with or without this skill, while\n  the recorded failures were all main sessions linting rows they had just\n  written. The isolation is the lever; do not skip it because L1 looked clean.\n  Models also under-ask unless pushed to, which is why the brief demands\n  its questions.\n- Every finding is either fixed in the row with evidence, or answered in the\n  row. A divergence between its restated goal and yours is a defect in the row.\n- Post its report on the row as a comment, with the fixes it caused.\n\nMeasured cost: about 81k tokens and 5 minutes per row. Spend it only on rows\nthat passed L1.\n\n### Promote, or do not\n\nPromote to Todo only when L0 is clean, every L1 claim is resolved, and the L2\nfindings are closed. A row with an open question stays in Backlog with the\nquestion written on it, ranked by what it changes. Ask only when the answer\nchanges the acceptance.\n\n## What to report\n\nReport per row, as a count and a pointer, never as \"groomed\" or \"ready\" alone:\n\n- rows promoted, and for each the L2 findings that were fixed;\n- rows left in Backlog, and the open question that holds each;\n- premises refuted, with the row and the evidence.\n\n## References\n\n- `skills\u002Fgroom\u002Freferences\u002Fvalidator-brief.md`: the L2 brief and reply shape.\n- `skills\u002Fgroom\u002Freferences\u002Fdefect-classes.md`: the measured classes, one line each,\n  with the row that measured it.\n- The evidence behind all of this is the Linear document \"Grooming evidence\n  dossier — what a groom must force that a linter cannot\", on CLOUD-1526.\n","# Grooming a row to Ready\n\nA Ready linter answers \"is the block well-formed\". It never answers \"is the row\ntrue\". Every defect this repository has measured in a groomed row was\n**well-formed and false**: a symbol that does not exist, a store census never\ntaken, a count with no command, a mechanism deferred behind a gate name, a test\nfile picked off a glob, a blocker that was never a blocker. So the linter's\nexit 0 is where grooming starts, never where it ends.\n\nThe deliverable of a grooming session is the **board**. A grooming session\nnever claims, readies or lands (CLOUD-1568).\n\n## The ladder — run it per row, cheapest rung first\n\nCopy this checklist and tick it per row. A row that fails a rung is fixed\nbefore the next, costlier rung is spent on it.\n\n```\nRow CLOUD-N:\n- [ ] L0 floor: `batten ready lint --issue CLOUD-N` findings fixed\n- [ ] L1 claims: every pointer, number and board claim resolved (below)\n- [ ] L1 decision: mechanism decided, tests discriminate, prose agrees with §7\n- [ ] L2 cold read: isolated validator report, every finding fixed or answered\n- [ ] promote, or leave in Backlog with the open question written on the row\n```\n\n### L0 — the floor\n\nRead the row through the verb (`batten mcp call Linear get_issue '{\"id\":\"CLOUD-N\"}'`),\nthen run `batten ready lint --issue CLOUD-N` and fix what it reports. Its exit 0\nmeans \"no known shape defect\". Never report it as \"ready\".\n\n### L1 — resolve every claim, in this session\n\nList every claim the row makes, then resolve each with the instrument\n`rules\u002Fscanning.md` names for its class. Do not reason from memory; a claim you\ndid not look up is unresolved.\n\n1. **Pointers.** Every path, `path::symbol`, rule id, task name, `input.*` key,\n   test name. Does it exist at `origin\u002Fmain`, and does it do what the row says?\n   A symbol that exists but does something else is a false premise.\n2. **Contents and live behaviour.** A claim about what a store, a session or a\n   queue CONTAINS is answered by counting it, not by reading the writer\n   (CLOUD-1523: correct in code, zero live instances).\n3. **Numbers.** Every figure names the command or fixture that reproduces it.\n   Re-run it. A number you cannot reproduce is removed or re-measured\n   (CLOUD-1522's 19,348 was a double count).\n4. **Board claims.** Blockers, \"decided on\", \"duplicate of\", a status. Search\n   the board before trusting any of them (`rules\u002Fscanning.md` row four). Look\n   for a refutation that exists elsewhere and was never written to this row.\n\nThen the decision checks, which are about the row against itself:\n\n5. **Decided, not deferred.** \"Needs a mechanism\", \"to be designed\", \"a gate\n   will decide\" beside a well-formed block is a deferral wearing a status\n   (CLOUD-1567). If the evidence in hand decides it, decide it in the row. If it\n   genuinely does not, write the open question and leave the row in Backlog.\n6. **Tests that discriminate.** Each `tests[]` entry names a real or clearly\n   new file, a case that fails without the change and passes with it, and a\n   `MUTANT \u003Cslug>|` row that `batten mutate` can reach or an honest statement\n   that it cannot (CLOUD-1526 clause 8). A file picked because it exists is the\n   defect `rules\u002Fscanning.md` measured, not a test obligation.\n7. **Prescription vs acceptance.** Does doing exactly what the row prescribes\n   satisfy the row's own §2 and §7? Where they disagree, the acceptance is the\n   spec (`mem:workflow\u002Fboard-states`).\n\nFix everything you can evidence, directly in the row: Backlog and Todo bodies\nare specs, and correcting them in place is what grooming is.\n\n### L2 — the isolated cold read\n\nThe implementer will start from nothing but the row, AGENTS.md and the repo.\nReproduce that: spawn one read-only `Explore` subagent per row, with the brief\nin `skills\u002Fgroom\u002Freferences\u002Fvalidator-brief.md` and the row body written to a\nscratch file it can Read. Rows are independent, so validators for several rows\nrun in parallel.\n\n- The validator is **report-only**. It never writes a file or the tracker. The\n  grooming session applies every fix itself, one row at a time.\n- Its value is that it did not write the row. Measured on CLOUD-1968: a fresh\n  isolated read caught every seeded defect with or without this skill, while\n  the recorded failures were all main sessions linting rows they had just\n  written. The isolation is the lever; do not skip it because L1 looked clean.\n  Models also under-ask unless pushed to, which is why the brief demands\n  its questions.\n- Every finding is either fixed in the row with evidence, or answered in the\n  row. A divergence between its restated goal and yours is a defect in the row.\n- Post its report on the row as a comment, with the fixes it caused.\n\nMeasured cost: about 81k tokens and 5 minutes per row. Spend it only on rows\nthat passed L1.\n\n### Promote, or do not\n\nPromote to Todo only when L0 is clean, every L1 claim is resolved, and the L2\nfindings are closed. A row with an open question stays in Backlog with the\nquestion written on it, ranked by what it changes. Ask only when the answer\nchanges the acceptance.\n\n## What to report\n\nReport per row, as a count and a pointer, never as \"groomed\" or \"ready\" alone:\n\n- rows promoted, and for each the L2 findings that were fixed;\n- rows left in Backlog, and the open question that holds each;\n- premises refuted, with the row and the evidence.\n\n## References\n\n- `skills\u002Fgroom\u002Freferences\u002Fvalidator-brief.md`: the L2 brief and reply shape.\n- `skills\u002Fgroom\u002Freferences\u002Fdefect-classes.md`: the measured classes, one line each,\n  with the row that measured it.\n- The evidence behind all of this is the Linear document \"Grooming evidence\n  dossier — what a groom must force that a linter cannot\", on CLOUD-1526.\n",{"name":5,"description":7},"main","fa67fcdbe97054cb5ef7dec5b12707702744229e","skills\u002Fgroom",0,"https:\u002F\u002Fgithub.com\u002Fbutton-inc\u002Fbatten","https:\u002F\u002Fgithub.com\u002Fbutton-inc","https:\u002F\u002Fbutton.is","2026-11-15 05:32:52",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add button-inc\u002Fbatten --skill groom",{"tool":38,"command":39},"bunx","bunx skills add button-inc\u002Fbatten --skill groom",{"tool":41,"command":42},"pnpm","pnpm dlx skills add button-inc\u002Fbatten --skill groom","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fbutton-inc\u002Fbatten\u002Fblob\u002Fmain\u002Fskills\u002Fgroom\u002FSKILL.md",true,[46,52,56],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"evals\u002Fevals.json","REFERENCE","参考文档",6265,"https:\u002F\u002Fgithub.com\u002Fbutton-inc\u002Fbatten\u002Fblob\u002Fmain\u002Fskills\u002Fgroom\u002Fevals\u002Fevals.json",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Fdefect-classes.md",2417,"https:\u002F\u002Fgithub.com\u002Fbutton-inc\u002Fbatten\u002Fblob\u002Fmain\u002Fskills\u002Fgroom\u002Freferences\u002Fdefect-classes.md",{"path":57,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":58,"sourceHtmlUrl":59},"references\u002Fvalidator-brief.md",2125,"https:\u002F\u002Fgithub.com\u002Fbutton-inc\u002Fbatten\u002Fblob\u002Fmain\u002Fskills\u002Fgroom\u002Freferences\u002Fvalidator-brief.md",[61,68,75,80,85],{"skillNo":62,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":65,"categories":66,"keywords":67,"publishedAt":19},"01M7A7REKD04W1H02ACWZH3XPY","serena","Choosing the instrument for a code question — when a symbol tool answers something a text scanner cannot, and which of Read, Grep, Glob or Serena is the right reach. Use when navigating or editing code you have not already read, when about to search the tree, or when a shell text utility is about to be pointed at a repository path.","https:\u002F\u002Fgithub.com\u002Fbutton-inc\u002Fbatten\u002Fblob\u002Fmain\u002Fskills\u002Fserena\u002FSKILL.md",[],[],{"skillNo":69,"name":70,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":72,"categories":73,"keywords":74,"publishedAt":19},"01M7A7R1Z3DYMGKT7PEEA2SK4P","batten","How to work in a repository gated by Batten — when to consult `batten check` versus trust the hook, how to read a pointer-only finding without re-running anything, what a deny is telling you, and how to read a verification receipt. Use when a Batten hook denies a call, when a `batten` command reports a finding, or before claiming work is done in a repository that has a batten.toml.","https:\u002F\u002Fgithub.com\u002Fbutton-inc\u002Fbatten\u002Fblob\u002Fmain\u002Fskills\u002Fbatten\u002FSKILL.md",[],[],{"skillNo":76,"name":63,"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":77,"categories":78,"keywords":79,"publishedAt":19},"01M7A7R10WCEPF0Q2QMTKVG5BR","https:\u002F\u002Fgithub.com\u002Fbutton-inc\u002Fbatten\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fserena\u002FSKILL.md",[],[],{"skillNo":81,"name":5,"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":82,"categories":83,"keywords":84,"publishedAt":19},"01M7A7QYB00DEWZZKZS7M18QBT","https:\u002F\u002Fgithub.com\u002Fbutton-inc\u002Fbatten\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fgroom\u002FSKILL.md",[],[],{"skillNo":86,"name":70,"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":87,"categories":88,"keywords":89,"publishedAt":19},"01M7A7QX3PZ6BD75D18QZB5AWG","https:\u002F\u002Fgithub.com\u002Fbutton-inc\u002Fbatten\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fbatten\u002FSKILL.md",[],[]]