[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CZ49JNN2QPHBPTWTP4NPF9":3,"skill-same:01M7CZ49JNN2QPHBPTWTP4NPF9":60},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"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":32,"commitSha":33,"skillDirectory":34,"forksCount":35,"repositoryHtmlUrl":30,"repositoryHomepageUrl":6,"ownerHtmlUrl":36,"ownerBlogUrl":37,"ownerDescription":38,"updatedAt":39,"install":40,"resources":53},"01M7CZ49JNN2QPHBPTWTP4NPF9","padosoft-agent-instructions-sync",null,"Use this skill whenever a file that instructs a coding agent is created, changed or deleted — CLAUDE.md, AGENTS.md, a rule or skill folder, a Copilot instructions file, a Gemini or Cursor or Codex configuration — in a repository where more than one agent is used. Also when the user says one agent follows a rule the others ignore, that a convention was applied by one tool and not another, that an instructions file was truncated, or asks how to keep the instruction files aligned. It gives the one-source-many-targets contract, the per-target limits that force a transformation, and the checks that catch a target left behind. Do not use it to write the content of a rule, nor to build a skill for publication (padosoft-skill-creator covers that).","padosoft\u002Fskills","padosoft","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F10467753?v=4","01M7CZ3FRSM851DK225E3KK2VF","01M7CZ3FSY016E21VS6EABM21V",1,"MIT","https:\u002F\u002Fgithub.com\u002Fpadosoft\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fpadosoft-agent-instructions-sync\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: padosoft-agent-instructions-sync\ndescription: >-\n  Use this skill whenever a file that instructs a coding agent is created, changed or deleted — CLAUDE.md,\n  AGENTS.md, a rule or skill folder, a Copilot instructions file, a Gemini or Cursor or Codex configuration\n  — in a repository where more than one agent is used. Also when the user says one agent follows a rule the\n  others ignore, that a convention was applied by one tool and not another, that an instructions file was\n  truncated, or asks how to keep the instruction files aligned. It gives the one-source-many-targets\n  contract, the per-target limits that force a transformation, and the checks that catch a target left\n  behind. Do not use it to write the content of a rule, nor to build a skill for publication\n  (padosoft-skill-creator covers that).\nlicense: MIT\ncompatibility: >-\n  Any repository with instruction files for more than one coding agent. The target-specific constraints are\n  stated as classes rather than product names, because they change faster than the rule does.\nmetadata:\n  version: 0.1.0\n  author: Padosoft\n  summary: A rule one agent knows and the others do not is worse than no rule.\n  profiles: core\n  scope: global\n  repository: https:\u002F\u002Fgithub.com\u002Fpadosoft\u002Fskills\n  keywords: claude.md, agents.md, copilot, gemini, cursor, instructions, rules, sync, multi-agent\n---\n\n# Agent instructions sync\n\nA repository used with more than one coding agent has the same rules written in several places, in formats\nthat are not interchangeable. Change one and the others keep instructing the old behaviour — silently,\nbecause nothing in a repository fails when two instruction files disagree.\n\n**The rule: one authoritative source, every other file derived from it in the same change. A rule one agent\nknows and the others do not is worse than no rule, because the disagreement gets attributed to the code.**\n\n---\n\n## 1. Declare the source and the targets, once\n\nWrite the mapping down in the repository, next to the source. It is the contract, and it is the thing that\ntells the next person what they are about to leave behind.\n\n| Source (authoritative) | Target A | Target B |\n|---|---|---|\n| the per-topic rule files | one instruction file each, path-scoped | one section each, appended to a single file |\n| the top-level project instructions | the tool's own top-level instructions file | a section in the same single file |\n| skills, packaged workflows | not supported — inline the critical part | not supported — inline the critical part |\n| subagent definitions | not supported — record the gap | not supported — record the gap |\n\nTwo things this table is for: telling you which targets exist at all, and **making the unsupported cells\nvisible**. A capability that one agent has and another does not is a gap you accept on purpose, not one you\ndiscover when the second agent does the wrong thing.\n\n## 2. The targets are not copies. They are transformations\n\nEach target imposes constraints that change the text, and the constraints are the reason a copy-paste\napproach quietly fails.\n\n- **A size limit that truncates.** Some review tools read only the first few thousand characters of an\n  instruction file and silently ignore the rest — so the rule at the bottom does not exist. Check the size\n  as part of the change, and condense rather than letting it be cut.\n- **Path scoping, or none.** One tool applies a file only to matching paths; another applies everything to\n  everything. When the target has no scoping, the instruction has to carry its own applicability in the\n  text, or it fires where it does not belong.\n- **Language.** Pick one language for all instruction files and keep it. Mixed-language instructions are\n  read worse by every tool, and they make a diff between source and target unreadable for a human.\n- **Condensation is allowed. Contradiction is not.** A target may drop examples, merge sections and lose\n  detail. It must never state a different rule from the source — that is how two agents end up enforcing\n  opposite conventions on the same file.\n\n## 3. Deletion propagates too\n\nThe failure nobody plans for: a rule is removed from the source and left in the targets. An agent then\nenforces a convention the team has abandoned, and the person arguing with it has nothing to point at\nbecause the source no longer mentions it.\n\nCreate, change, **delete** — all three cross the whole mapping.\n\n## 4. Make the gap detectable\n\nA mapping table maintained by hand rots like any other document. Add whichever of these fits the repository:\n\n- **A check that every source file has its declared targets**, and that no target exists without a source —\n  the second half catches the orphan left by a deletion.\n- **A size check per target**, failing before the truncating tool silently drops half the file.\n- **A marker linking each target back to its source** — a stable identifier in both — so a reader of the\n  target can find the authority in one search.\n\nNone of this needs a framework. It needs the mapping to exist in a form a script can read: see\n**`padosoft-ci-workflow-gates`** for why a manifest that does not enumerate itself stays green while half\nthe contract is deleted.\n\n## 5. The same discipline for what the instructions quote\n\nAn instruction file that names a helper, a path, a command or a flag is documentation, and it rots exactly\nlike documentation — with the difference that the agent follows the stale instruction **confidently**, which\nis worse than having no instruction. See **`padosoft-docs-match-code`**, and treat a rename as a change to\nevery file that mentions the old name — **`padosoft-contract-changes`** is the same procedure at a different\naltitude.\n\n---\n\n## Gotchas\n\n- **Nothing fails when two instruction files disagree.** There is no test, no build, no runtime. The only\n  symptom is two agents behaving differently, which gets blamed on the models.\n- **The truncated file is the dangerous one**, because it looks present and complete in the editor.\n- **\"I will sync the others later\" means the others are now wrong**, and the person who next reads them has\n  no way to know which one is current.\n- **A target with no path scoping applies your backend rule to the frontend.** Either say so in the text, or\n  accept that the rule fires out of place.\n- **Do not let the derived file become the one people edit.** If a target is easier to open than the source,\n  it will be edited there, and the next sync silently reverts it. Say in the target that it is generated.\n\n## Checklist\n\n- [ ] The mapping between source and every target is written down in the repository\n- [ ] Unsupported capabilities are listed as gaps, not left implicit\n- [ ] Every create, change **and** delete crossed the whole mapping in the same change\n- [ ] Each target respects its size limit, checked rather than assumed\n- [ ] Applicability is carried in the text wherever the target has no path scoping\n- [ ] One language across all instruction files\n- [ ] No target contradicts the source; condensation only\n- [ ] Targets say they are generated, and name their source\n- [ ] A check exists for missing targets and for orphaned ones\n\n## Final report\n\n```\nChange: \u003Cwhat was created \u002F modified \u002F deleted, in the source>\nTargets updated: \u003Cfile> (\u003Csize> \u002F \u003Climit>) · \u003Cfile> · …\nUnsupported: \u003Ccapability> → \u003Chow it was handled, or the accepted gap>\nDeletions propagated: yes | no | not applicable\nContradictions checked: \u003Cwhich rules were compared>\nDetection: \u003Cthe check that will catch the next missed target>\n```\n","# Agent instructions sync\n\nA repository used with more than one coding agent has the same rules written in several places, in formats\nthat are not interchangeable. Change one and the others keep instructing the old behaviour — silently,\nbecause nothing in a repository fails when two instruction files disagree.\n\n**The rule: one authoritative source, every other file derived from it in the same change. A rule one agent\nknows and the others do not is worse than no rule, because the disagreement gets attributed to the code.**\n\n---\n\n## 1. Declare the source and the targets, once\n\nWrite the mapping down in the repository, next to the source. It is the contract, and it is the thing that\ntells the next person what they are about to leave behind.\n\n| Source (authoritative) | Target A | Target B |\n|---|---|---|\n| the per-topic rule files | one instruction file each, path-scoped | one section each, appended to a single file |\n| the top-level project instructions | the tool's own top-level instructions file | a section in the same single file |\n| skills, packaged workflows | not supported — inline the critical part | not supported — inline the critical part |\n| subagent definitions | not supported — record the gap | not supported — record the gap |\n\nTwo things this table is for: telling you which targets exist at all, and **making the unsupported cells\nvisible**. A capability that one agent has and another does not is a gap you accept on purpose, not one you\ndiscover when the second agent does the wrong thing.\n\n## 2. The targets are not copies. They are transformations\n\nEach target imposes constraints that change the text, and the constraints are the reason a copy-paste\napproach quietly fails.\n\n- **A size limit that truncates.** Some review tools read only the first few thousand characters of an\n  instruction file and silently ignore the rest — so the rule at the bottom does not exist. Check the size\n  as part of the change, and condense rather than letting it be cut.\n- **Path scoping, or none.** One tool applies a file only to matching paths; another applies everything to\n  everything. When the target has no scoping, the instruction has to carry its own applicability in the\n  text, or it fires where it does not belong.\n- **Language.** Pick one language for all instruction files and keep it. Mixed-language instructions are\n  read worse by every tool, and they make a diff between source and target unreadable for a human.\n- **Condensation is allowed. Contradiction is not.** A target may drop examples, merge sections and lose\n  detail. It must never state a different rule from the source — that is how two agents end up enforcing\n  opposite conventions on the same file.\n\n## 3. Deletion propagates too\n\nThe failure nobody plans for: a rule is removed from the source and left in the targets. An agent then\nenforces a convention the team has abandoned, and the person arguing with it has nothing to point at\nbecause the source no longer mentions it.\n\nCreate, change, **delete** — all three cross the whole mapping.\n\n## 4. Make the gap detectable\n\nA mapping table maintained by hand rots like any other document. Add whichever of these fits the repository:\n\n- **A check that every source file has its declared targets**, and that no target exists without a source —\n  the second half catches the orphan left by a deletion.\n- **A size check per target**, failing before the truncating tool silently drops half the file.\n- **A marker linking each target back to its source** — a stable identifier in both — so a reader of the\n  target can find the authority in one search.\n\nNone of this needs a framework. It needs the mapping to exist in a form a script can read: see\n**`padosoft-ci-workflow-gates`** for why a manifest that does not enumerate itself stays green while half\nthe contract is deleted.\n\n## 5. The same discipline for what the instructions quote\n\nAn instruction file that names a helper, a path, a command or a flag is documentation, and it rots exactly\nlike documentation — with the difference that the agent follows the stale instruction **confidently**, which\nis worse than having no instruction. See **`padosoft-docs-match-code`**, and treat a rename as a change to\nevery file that mentions the old name — **`padosoft-contract-changes`** is the same procedure at a different\naltitude.\n\n---\n\n## Gotchas\n\n- **Nothing fails when two instruction files disagree.** There is no test, no build, no runtime. The only\n  symptom is two agents behaving differently, which gets blamed on the models.\n- **The truncated file is the dangerous one**, because it looks present and complete in the editor.\n- **\"I will sync the others later\" means the others are now wrong**, and the person who next reads them has\n  no way to know which one is current.\n- **A target with no path scoping applies your backend rule to the frontend.** Either say so in the text, or\n  accept that the rule fires out of place.\n- **Do not let the derived file become the one people edit.** If a target is easier to open than the source,\n  it will be edited there, and the next sync silently reverts it. Say in the target that it is generated.\n\n## Checklist\n\n- [ ] The mapping between source and every target is written down in the repository\n- [ ] Unsupported capabilities are listed as gaps, not left implicit\n- [ ] Every create, change **and** delete crossed the whole mapping in the same change\n- [ ] Each target respects its size limit, checked rather than assumed\n- [ ] Applicability is carried in the text wherever the target has no path scoping\n- [ ] One language across all instruction files\n- [ ] No target contradicts the source; condensation only\n- [ ] Targets say they are generated, and name their source\n- [ ] A check exists for missing targets and for orphaned ones\n\n## Final report\n\n```\nChange: \u003Cwhat was created \u002F modified \u002F deleted, in the source>\nTargets updated: \u003Cfile> (\u003Csize> \u002F \u003Climit>) · \u003Cfile> · …\nUnsupported: \u003Ccapability> → \u003Chow it was handled, or the accepted gap>\nDeletions propagated: yes | no | not applicable\nContradictions checked: \u003Cwhich rules were compared>\nDetection: \u003Cthe check that will catch the next missed target>\n```\n",{"name":5,"license":14,"metadata":23,"description":7,"compatibility":31},{"scope":24,"author":25,"summary":26,"version":27,"keywords":28,"profiles":29,"repository":30},"global","Padosoft","A rule one agent knows and the others do not is worse than no rule.","0.1.0","claude.md, agents.md, copilot, gemini, cursor, instructions, rules, sync, multi-agent","core","https:\u002F\u002Fgithub.com\u002Fpadosoft\u002Fskills","Any repository with instruction files for more than one coding agent. The target-specific constraints are stated as classes rather than product names, because they change faster than the rule does.","main","8ff266762b08cab7f5be5551fa221309dc9db48b","skills\u002Fpadosoft-agent-instructions-sync",0,"https:\u002F\u002Fgithub.com\u002Fpadosoft","https:\u002F\u002Fwww.padosoft.com","Software House","2026-10-19 01:02:17",{"source":8,"skillName":5,"commands":41,"prompt":51,"zipAvailable":52},[42,45,48],{"tool":43,"command":44},"npx","npx skills add padosoft\u002Fskills --skill padosoft-agent-instructions-sync",{"tool":46,"command":47},"bunx","bunx skills add padosoft\u002Fskills --skill padosoft-agent-instructions-sync",{"tool":49,"command":50},"pnpm","pnpm dlx skills add padosoft\u002Fskills --skill padosoft-agent-instructions-sync","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fpadosoft\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fpadosoft-agent-instructions-sync\u002FSKILL.md",true,[54],{"path":55,"resourceType":56,"resourceTypeDesc":57,"sizeBytes":58,"sourceHtmlUrl":59},"evals\u002Fqueries.json","REFERENCE","参考文档",1538,"https:\u002F\u002Fgithub.com\u002Fpadosoft\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fpadosoft-agent-instructions-sync\u002Fevals\u002Fqueries.json",[61,68,74,81,88,94],{"skillNo":62,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":65,"categories":66,"keywords":67,"publishedAt":18},"01M7CZNKCTMWK7F2J6KKPAR8YM","report-opus","Writes the weekly status report from the team's tickets. Use when the user asks for the weekly report or a status summary. Do not use it for incident reports.","https:\u002F\u002Fgithub.com\u002Fpadosoft\u002Fskills\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fskill-audit\u002Fproject\u002Fdot-claude\u002Fskills\u002Freport-opus\u002FSKILL.md",[],[],{"skillNo":69,"name":70,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":18},"01M7CZNFEN8XK6S8NC6KGREYGS","report-haiku","https:\u002F\u002Fgithub.com\u002Fpadosoft\u002Fskills\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fskill-audit\u002Fproject\u002Fdot-claude\u002Fskills\u002Freport-haiku\u002FSKILL.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":18},"01M7CZND182TKGGM7R4134ND00","kitt-stub","Builds the guided-tour reference for an admin page. Use when the user asks for the tour reference. Trigger \"tour\", \"reference\", \"guided tour\".","https:\u002F\u002Fgithub.com\u002Fpadosoft\u002Fskills\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fskill-audit\u002Fproject\u002Fdot-claude\u002Fskills\u002Fkitt-stub\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":18},"01M7CZN9F2A6BFTFW5RSRNAGB1","good-skill","Generates the release notes of a project from its merged pull requests. Use this skill when the user asks for release notes, a changelog entry or a summary of what shipped since the last tag. Do not use it to write commit messages, that is the job of a commit skill.","https:\u002F\u002Fgithub.com\u002Fpadosoft\u002Fskills\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fskill-audit\u002Fproject\u002Fdot-claude\u002Fskills\u002Fgood-skill\u002FSKILL.md",[],[],{"skillNo":89,"name":90,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":93,"publishedAt":18},"01M7CZN6BJASPD4DH8DFR63GQF","broken-frontmatter","https:\u002F\u002Fgithub.com\u002Fpadosoft\u002Fskills\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fskill-audit\u002Fproject\u002Fdot-claude\u002Fskills\u002Fbroken-frontmatter\u002FSKILL.md",[],[],{"skillNo":95,"name":96,"nameZh":6,"description":97,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":98,"categories":99,"keywords":100,"publishedAt":18},"01M7CZN498TRNC30GSYMMAZNXM","Bad_Skill","I can help you with reports.","https:\u002F\u002Fgithub.com\u002Fpadosoft\u002Fskills\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fskill-audit\u002Fproject\u002Fdot-claude\u002Fskills\u002Fbad-skill\u002FSKILL.md",[],[]]