[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CQ089EPPRFF22HA4QRFV5T":3,"skill-same:01M7CQ089EPPRFF22HA4QRFV5T":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":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M7CQ089EPPRFF22HA4QRFV5T","review",null,"Dispatch Canon's independent reviewer subagents for this branch and read their verdicts back from canon_review. Use once a change is built and verified, before shipping. This skill only dispatches and reads back -- the review itself runs inside the subagent, in a fresh context, and never in this session.","urban233\u002FCanon","urban233","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F81388165?v=4","01M7CPZG1XX0BZ1KTRAJWTGWFH","Martin Urban","01M7CPZG2JPZEX3EJEG98XGTA7",1,"BSD-3-Clause","https:\u002F\u002Fgithub.com\u002Furban233\u002FCanon\u002Fblob\u002Fmain\u002Fplugins\u002Fantigravity\u002Fskills\u002Freview\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: review\ndescription: Dispatch Canon's independent reviewer subagents for this branch and read their verdicts back from canon_review. Use once a change is built and verified, before shipping. This skill only dispatches and reads back -- the review itself runs inside the subagent, in a fresh context, and never in this session.\n---\n\n# Review\n\nInvariant III is that nothing ships on the agent's own word, and a review\nyou perform yourself is your own word. So this skill's whole job is to\nhand the change to something that did not write it, and to read back what\nthat something actually said.\n\n## Dispatch\n\n1. **Call `canon_review`.** It reports `reviewers_called_for` -- always\n   `reviewer`, plus `risk-reviewer` when the changed paths touch a risk\n   surface. You do not choose them and there is no waiver to write; the\n   diff decides.\n2. **Dispatch each reviewer it names**, giving it the `base..HEAD` range,\n   the saved plan and its `## Non-goals`, and the evidence that checks\n   ran. Never your own reasoning about the change -- a reviewer's value\n   comes precisely from not having seen it.\n\n   Antigravity bundles both reviewers with this plugin, under `agents\u002F`,\n   so `invoke_subagent` can reach them by name once the plugin is\n   enabled. If neither resolves there is nothing to dispatch, and the\n   honest answer is to say so -- not to review the change yourself,\n   which is the one thing this skill exists to prevent.\n3. **Read the verdict out of the reviewer's own final message.**\n\n   On Claude Code and Codex a `SubagentStop` hook captures that message\n   and `canon_review` reads it back, so the verdict reaching `ship` is\n   the reviewer's own words rather than this session's retelling.\n   **Antigravity fires no such event**, so that capture does not happen\n   here and `canon_review` will report no stored verdict. Quote the\n   reviewer's closing verdict line verbatim instead of paraphrasing it,\n   and say that it was read in-session. Do not present a paraphrase as\n   though it came back through `canon_review`: a weakened guarantee\n   stated plainly is worth more than a strong one that isn't true.\n   See this plugin's README for the standing limitation.\n\n## When a reviewer has seen this branch before\n\n`canon_review`'s `verdicts` names, per reviewer, the `head` that reviewer\nlast looked at and whether that is `stale`. When a reviewer's own verdict\nis stale, **tell it which commit it last reviewed** and ask it to review\n`\u003Chead>..HEAD` as a delta **in addition to** the full `base..HEAD` range,\nnever instead of it.\n\nThis matters because the most expensive failure in a repair loop is a fix\nthat closes the reported finding and quietly breaks something the reviewer\nalready passed. Re-reading the whole diff makes that regression look like\npart of the change; reading the delta against a commit already approved is\nwhat makes it visible.\n\n**After a READY verdict, confirm the delta only.** When a reviewer's own\nlast verdict is READY FOR HUMAN APPROVAL and its `head` is an ancestor of\nHEAD -- you only closed its notes, or touched docs -- it has already\npassed everything up to that commit. Ask it to review `\u003Chead>..HEAD`\nalone and return a fresh verdict. The full range is for repair after\nCHANGES REQUIRED, where a fix can break what was passed; after READY\nthere is nothing unpassed outside the delta, and re-reading the whole\nrange costs a full round for no new coverage. The verdict is still fresh\nat HEAD, so `canon_ship` is satisfied the same way.\n\n**Read `stale` per reviewer, not from the combined `verdict`.** The\ncombined `stale` is true when *any* reviewer's verdict is stale, while the\ncombined `head` belongs to whichever verdict ranked worst. With two\nreviewers in play those two fields can describe different agents, and a\ndelta computed from the combined pair can span the wrong commits entirely.\n`verdicts[\u003Cname>].head` and `verdicts[\u003Cname>].stale` always describe the\nsame reviewer.\n\n## Stop after two rounds on the same change\n\nA third dispatch for the same change is the signal that the problem is not\nthe code. When you have already addressed two `CHANGES REQUIRED` verdicts\non this branch, **stop and ask the developer** rather than dispatching\nagain. Grinding is the most expensive failure mode there is, and the\npoint at which to spend a human's attention is before the third attempt,\nnot after the fifth.\n\nThis is imposed on you, not on the developer: it is not a gate, it blocks\nnothing, and `canon_ship` never reads it. Nothing counts the rounds for\nyou either, deliberately -- a counter in the repository is\n`round-state.json`, and a counter in session state is both unreadable by\n`canon_ship` and lost on the next session anyway. So this is yours to\nhold, from the rounds in this conversation. The honest consequence: after\na compaction the count is gone, and the rule degrades to judgement. On\nClaude Code and Codex, `canon_review` counts the rounds from captured\nverdicts; Antigravity captures none (step 3 above), so here the count\nreally is only in this conversation.\n","# Review\n\nInvariant III is that nothing ships on the agent's own word, and a review\nyou perform yourself is your own word. So this skill's whole job is to\nhand the change to something that did not write it, and to read back what\nthat something actually said.\n\n## Dispatch\n\n1. **Call `canon_review`.** It reports `reviewers_called_for` -- always\n   `reviewer`, plus `risk-reviewer` when the changed paths touch a risk\n   surface. You do not choose them and there is no waiver to write; the\n   diff decides.\n2. **Dispatch each reviewer it names**, giving it the `base..HEAD` range,\n   the saved plan and its `## Non-goals`, and the evidence that checks\n   ran. Never your own reasoning about the change -- a reviewer's value\n   comes precisely from not having seen it.\n\n   Antigravity bundles both reviewers with this plugin, under `agents\u002F`,\n   so `invoke_subagent` can reach them by name once the plugin is\n   enabled. If neither resolves there is nothing to dispatch, and the\n   honest answer is to say so -- not to review the change yourself,\n   which is the one thing this skill exists to prevent.\n3. **Read the verdict out of the reviewer's own final message.**\n\n   On Claude Code and Codex a `SubagentStop` hook captures that message\n   and `canon_review` reads it back, so the verdict reaching `ship` is\n   the reviewer's own words rather than this session's retelling.\n   **Antigravity fires no such event**, so that capture does not happen\n   here and `canon_review` will report no stored verdict. Quote the\n   reviewer's closing verdict line verbatim instead of paraphrasing it,\n   and say that it was read in-session. Do not present a paraphrase as\n   though it came back through `canon_review`: a weakened guarantee\n   stated plainly is worth more than a strong one that isn't true.\n   See this plugin's README for the standing limitation.\n\n## When a reviewer has seen this branch before\n\n`canon_review`'s `verdicts` names, per reviewer, the `head` that reviewer\nlast looked at and whether that is `stale`. When a reviewer's own verdict\nis stale, **tell it which commit it last reviewed** and ask it to review\n`\u003Chead>..HEAD` as a delta **in addition to** the full `base..HEAD` range,\nnever instead of it.\n\nThis matters because the most expensive failure in a repair loop is a fix\nthat closes the reported finding and quietly breaks something the reviewer\nalready passed. Re-reading the whole diff makes that regression look like\npart of the change; reading the delta against a commit already approved is\nwhat makes it visible.\n\n**After a READY verdict, confirm the delta only.** When a reviewer's own\nlast verdict is READY FOR HUMAN APPROVAL and its `head` is an ancestor of\nHEAD -- you only closed its notes, or touched docs -- it has already\npassed everything up to that commit. Ask it to review `\u003Chead>..HEAD`\nalone and return a fresh verdict. The full range is for repair after\nCHANGES REQUIRED, where a fix can break what was passed; after READY\nthere is nothing unpassed outside the delta, and re-reading the whole\nrange costs a full round for no new coverage. The verdict is still fresh\nat HEAD, so `canon_ship` is satisfied the same way.\n\n**Read `stale` per reviewer, not from the combined `verdict`.** The\ncombined `stale` is true when *any* reviewer's verdict is stale, while the\ncombined `head` belongs to whichever verdict ranked worst. With two\nreviewers in play those two fields can describe different agents, and a\ndelta computed from the combined pair can span the wrong commits entirely.\n`verdicts[\u003Cname>].head` and `verdicts[\u003Cname>].stale` always describe the\nsame reviewer.\n\n## Stop after two rounds on the same change\n\nA third dispatch for the same change is the signal that the problem is not\nthe code. When you have already addressed two `CHANGES REQUIRED` verdicts\non this branch, **stop and ask the developer** rather than dispatching\nagain. Grinding is the most expensive failure mode there is, and the\npoint at which to spend a human's attention is before the third attempt,\nnot after the fifth.\n\nThis is imposed on you, not on the developer: it is not a gate, it blocks\nnothing, and `canon_ship` never reads it. Nothing counts the rounds for\nyou either, deliberately -- a counter in the repository is\n`round-state.json`, and a counter in session state is both unreadable by\n`canon_ship` and lost on the next session anyway. So this is yours to\nhold, from the rounds in this conversation. The honest consequence: after\na compaction the count is gone, and the rule degrades to judgement. On\nClaude Code and Codex, `canon_review` counts the rounds from captured\nverdicts; Antigravity captures none (step 3 above), so here the count\nreally is only in this conversation.\n",{"name":5,"description":7},"main","453e779ab8a331947f8cb2b1168cb0ba8f1f3256","plugins\u002Fantigravity\u002Fskills\u002Freview",0,"https:\u002F\u002Fgithub.com\u002Furban233\u002FCanon","https:\u002F\u002Fgithub.com\u002Furban233","2026-10-01 06:44:57",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add urban233\u002FCanon --skill review",{"tool":37,"command":38},"bunx","bunx skills add urban233\u002FCanon --skill review",{"tool":40,"command":41},"pnpm","pnpm dlx skills add urban233\u002FCanon --skill review","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Furban233\u002FCanon\u002Fblob\u002Fmain\u002Fplugins\u002Fantigravity\u002Fskills\u002Freview\u002FSKILL.md",true,[],[46,53,60,65,71,78],{"skillNo":47,"name":48,"nameZh":6,"description":49,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":50,"categories":51,"keywords":52,"publishedAt":19},"01M7CQ48XK94BYRMQVDCQRTY77","testing-craft","Provides Google-derived testing strategy, test-writing craft, and test-suite health reference, distilled from Software Engineering at Google (the Flamingo book), the Google Testing Blog, and Google's published testing-infrastructure research. Canon's `reviewer` subagent reads the writing-craft and strategy references as review criteria. Invoke directly to design a test strategy, audit an existing suite's health, or triage a flaky or brittle test.","https:\u002F\u002Fgithub.com\u002Furban233\u002FCanon\u002Fblob\u002Fmain\u002Fplugins\u002Fcodex\u002Fskills\u002Ftesting-craft\u002FSKILL.md",[],[],{"skillNo":54,"name":55,"nameZh":6,"description":56,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":57,"categories":58,"keywords":59,"publishedAt":19},"01M7CQ465824PE31REMF8HFB74","ship","Decide whether a change is ready to hand to a human, and open the pull request when it is. Use once the work is done and verified, before opening a PR or telling the developer the change is ready.","https:\u002F\u002Fgithub.com\u002Furban233\u002FCanon\u002Fblob\u002Fmain\u002Fplugins\u002Fcodex\u002Fskills\u002Fship\u002FSKILL.md",[],[],{"skillNo":61,"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":62,"categories":63,"keywords":64,"publishedAt":19},"01M7CQ40YM28KQ9H79HEC2T0SB","https:\u002F\u002Fgithub.com\u002Furban233\u002FCanon\u002Fblob\u002Fmain\u002Fplugins\u002Fcodex\u002Fskills\u002Freview\u002FSKILL.md",[],[],{"skillNo":66,"name":67,"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":68,"categories":69,"keywords":70,"publishedAt":19},"01M7CQ3ZK97QCHFZ7NNP6E1EZP","review-change","https:\u002F\u002Fgithub.com\u002Furban233\u002FCanon\u002Fblob\u002Fmain\u002Fplugins\u002Fcodex\u002Fskills\u002Freview-change\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},"01M7CQ3KTVG5BAW4ZVMNYS1AFE","plan","Shape what plan mode drafts, so the plan Canon saves needs no follow-up. Use whenever drafting a plan for approval in plan mode, before writing it to .canon\u002Fplans\u002F.","https:\u002F\u002Fgithub.com\u002Furban233\u002FCanon\u002Fblob\u002Fmain\u002Fplugins\u002Fcodex\u002Fskills\u002Fplan\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},"01M7CQ2XHVX3F1GT08CCCY225D","frame","Surface the open questions before planning a multi-branch feature, and shape the feature plan mode drafts. Use before entering plan mode for work that obviously won't fit in one branch and one pull request -- never for a change that already does.","https:\u002F\u002Fgithub.com\u002Furban233\u002FCanon\u002Fblob\u002Fmain\u002Fplugins\u002Fcodex\u002Fskills\u002Fframe\u002FSKILL.md",[],[]]