[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7D6JWDSP3JD275W67MXSATY":3,"skill-same:01M7D6JWDSP3JD275W67MXSATY":77},{"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":26,"commitSha":27,"skillDirectory":28,"forksCount":14,"repositoryHtmlUrl":29,"repositoryHomepageUrl":30,"ownerHtmlUrl":31,"ownerBlogUrl":6,"ownerDescription":32,"updatedAt":33,"install":34,"resources":46},"01M7D6JWDSP3JD275W67MXSATY","ce-retune",null,"Retune a skill corpus for a new model, measurement-first: mine the run archive for a baseline, establish a noise floor, audit the corpus adversarially, then cut in measured passes until a pre-registered bar clears. Requires a benchmark harness that can A\u002FB two builds of the corpus; refuses without one.","crdant\u002Fcompound-engineering-plugin","crdant","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F50028?v=4","01M7D6FTRNYFR3V2B7XDZZYH57","Chuck D'Antonio","01M7D6FTSCJWPR3R1T1AY123GD",0,"MIT","https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Fskills\u002Fce-retune\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: ce-retune\ndescription: \"Retune a skill corpus for a new model, measurement-first: mine the run archive for a baseline, establish a noise floor, audit the corpus adversarially, then cut in measured passes until a pre-registered bar clears. Requires a benchmark harness that can A\u002FB two builds of the corpus; refuses without one.\"\ndisable-model-invocation: true\nargument-hint: \"[target model or symptom] [path to the corpus, defaults to .\u002Fskills] [bar:\u003Cn> consecutive clean runs]\"\n---\n\n# Retune a Corpus for a New Model\n\nA corpus that degrades on a new model is a measurement problem before it is a writing problem: rewriting what looks wrong produces a plausible fix list and no way to know whether any item mattered.\n\n**Outcome:** a corpus whose measured behavior on the target model clears a bar registered before any change, with the regression classes removed and each removal attributable.\n\n**Done:** the bar is cleared, or the run reports the specific claim it could not support. A green test suite is not done: it proves nothing broke, not that behavior improved.\n\n**Non-goal:** word reduction. Leanness and performance are separate programs that share a corpus; only one of them is the result here. Report completion, not word count.\n\n\n## Phase 0: the measurement gate — check this first\n\nThis skill cannot run without a way to observe behavior. Check for all three, and name whichever is missing:\n\n1. **A run archive or a harness that produces one** — per-run logs carrying the tool-call trace, a terminal marker, token counts, and the final message.\n2. **A build selector** — the harness can point a run at a specific source checkout of the corpus (a `--plugin-dir`-style override, a configurable skills path, an env var), so two builds are comparable under one runner.\n3. **A repeatable task** the corpus actually executes end to end.\n\nIf any is missing, **stop and say so**, naming what to build. Do not fall back to a static audit and present it as retuning: an audit can say what looks cuttable and never whether cutting helped. An audit-only pass is a legitimate thing to want; it is a different request.\n\nState the target model and the harness you found before continuing.\n\n## The phases\n\nThey run in order, and each names the reference it cannot start without. Read `references\u002Fworkflow-shapes.md` before dispatching any phase: the wrong orchestration shape is the common failure. Fan out by disjoint file ownership, never by item. Items cross files, and agents that share a file lose each other's edits.\n\n1. **Mine the archive** before spending a run — `references\u002Fbaseline-mining.md`. Historical runs are a free baseline, usually larger than any experiment affordable now.\n2. **Establish the noise floor** — `references\u002Fnoise-floor.md`. Run the harness against **two identical copies** of the corpus, same commit on both sides; whatever difference appears is the floor every later claim must clear. **Register the bar now, in writing, before any change exists.** A bar chosen after seeing results is not a bar.\n3. **Audit the corpus adversarially** — `references\u002Fcorpus-audit.md`. One agent per skill proposes cuts; a second per skill does the opposite and defends the existing prose. **The two passes require independent contexts.** If the host exposes no way to run them as separate agents, report that as a blocker and stop the audit — do not argue both sides in one context and present the result as an audit.\n4. **Cut in surgical passes**, one problem per agent — `references\u002Fcut-passes.md`, and `references\u002Fhalt-taxonomy.md` when the symptom is stalling, halting, or a run that ends while naming work it did not do. Two rules bound every pass, whatever class it is cutting. **Never edit a test to make a suite green**: a removed string a test pins is a finding to report, not a test to weaken. And **not every stop is the enemy.** Some workflows exist to stop and ask; that is the product. Sort every stop by who is actually on the other side before touching it. `references\u002Fhalt-taxonomy.md` carries the screens that decide, so read them before cutting any stop.\n5. **Measure, then let the failure choose the next fix** — `references\u002Fcut-passes.md` again for what each failure site means and for auditing the phases the instrument never enters. Loop 4 and 5 until the registered bar clears. Then stop; a bar cleared is done. Also report what stayed unmeasured: a cleared bar never implies coverage it does not have.\n6. **Ship** — `references\u002Fcut-passes.md` carries what the commits and the write-up must preserve.\n","# Retune a Corpus for a New Model\n\nA corpus that degrades on a new model is a measurement problem before it is a writing problem: rewriting what looks wrong produces a plausible fix list and no way to know whether any item mattered.\n\n**Outcome:** a corpus whose measured behavior on the target model clears a bar registered before any change, with the regression classes removed and each removal attributable.\n\n**Done:** the bar is cleared, or the run reports the specific claim it could not support. A green test suite is not done: it proves nothing broke, not that behavior improved.\n\n**Non-goal:** word reduction. Leanness and performance are separate programs that share a corpus; only one of them is the result here. Report completion, not word count.\n\n\n## Phase 0: the measurement gate — check this first\n\nThis skill cannot run without a way to observe behavior. Check for all three, and name whichever is missing:\n\n1. **A run archive or a harness that produces one** — per-run logs carrying the tool-call trace, a terminal marker, token counts, and the final message.\n2. **A build selector** — the harness can point a run at a specific source checkout of the corpus (a `--plugin-dir`-style override, a configurable skills path, an env var), so two builds are comparable under one runner.\n3. **A repeatable task** the corpus actually executes end to end.\n\nIf any is missing, **stop and say so**, naming what to build. Do not fall back to a static audit and present it as retuning: an audit can say what looks cuttable and never whether cutting helped. An audit-only pass is a legitimate thing to want; it is a different request.\n\nState the target model and the harness you found before continuing.\n\n## The phases\n\nThey run in order, and each names the reference it cannot start without. Read `references\u002Fworkflow-shapes.md` before dispatching any phase: the wrong orchestration shape is the common failure. Fan out by disjoint file ownership, never by item. Items cross files, and agents that share a file lose each other's edits.\n\n1. **Mine the archive** before spending a run — `references\u002Fbaseline-mining.md`. Historical runs are a free baseline, usually larger than any experiment affordable now.\n2. **Establish the noise floor** — `references\u002Fnoise-floor.md`. Run the harness against **two identical copies** of the corpus, same commit on both sides; whatever difference appears is the floor every later claim must clear. **Register the bar now, in writing, before any change exists.** A bar chosen after seeing results is not a bar.\n3. **Audit the corpus adversarially** — `references\u002Fcorpus-audit.md`. One agent per skill proposes cuts; a second per skill does the opposite and defends the existing prose. **The two passes require independent contexts.** If the host exposes no way to run them as separate agents, report that as a blocker and stop the audit — do not argue both sides in one context and present the result as an audit.\n4. **Cut in surgical passes**, one problem per agent — `references\u002Fcut-passes.md`, and `references\u002Fhalt-taxonomy.md` when the symptom is stalling, halting, or a run that ends while naming work it did not do. Two rules bound every pass, whatever class it is cutting. **Never edit a test to make a suite green**: a removed string a test pins is a finding to report, not a test to weaken. And **not every stop is the enemy.** Some workflows exist to stop and ask; that is the product. Sort every stop by who is actually on the other side before touching it. `references\u002Fhalt-taxonomy.md` carries the screens that decide, so read them before cutting any stop.\n5. **Measure, then let the failure choose the next fix** — `references\u002Fcut-passes.md` again for what each failure site means and for auditing the phases the instrument never enters. Loop 4 and 5 until the registered bar clears. Then stop; a bar cleared is done. Also report what stayed unmeasured: a cleared bar never implies coverage it does not have.\n6. **Ship** — `references\u002Fcut-passes.md` carries what the commits and the write-up must preserve.\n",{"name":5,"description":7,"argument-hint":24,"disable-model-invocation":25},"[target model or symptom] [path to the corpus, defaults to .\u002Fskills] [bar:\u003Cn> consecutive clean runs]",true,"main","f794948b1838eb38f1f47e83176b85e2c9cb3873","skills\u002Fce-retune","https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin","https:\u002F\u002Fevery.to\u002Fguides\u002Fcompound-engineering","https:\u002F\u002Fgithub.com\u002Fcrdant","System architect & customer advocate. Father to 2(+1), open source & devops advocate, open water swimmer. Zealous cook, eater & cocktail composer.","2026-09-22 21:16:30",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":25},[36,39,42],{"tool":37,"command":38},"npx","npx skills add crdant\u002Fcompound-engineering-plugin --skill ce-retune",{"tool":40,"command":41},"bunx","bunx skills add crdant\u002Fcompound-engineering-plugin --skill ce-retune",{"tool":43,"command":44},"pnpm","pnpm dlx skills add crdant\u002Fcompound-engineering-plugin --skill ce-retune","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Fskills\u002Fce-retune\u002FSKILL.md",[47,53,57,61,65,69,73],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"agents\u002Fopenai.yaml","REFERENCE","参考文档",43,"https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Fskills\u002Fce-retune\u002Fagents\u002Fopenai.yaml",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Fbaseline-mining.md",10666,"https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Fskills\u002Fce-retune\u002Freferences\u002Fbaseline-mining.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Fcorpus-audit.md",12768,"https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Fskills\u002Fce-retune\u002Freferences\u002Fcorpus-audit.md",{"path":62,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":63,"sourceHtmlUrl":64},"references\u002Fcut-passes.md",10984,"https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Fskills\u002Fce-retune\u002Freferences\u002Fcut-passes.md",{"path":66,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":67,"sourceHtmlUrl":68},"references\u002Fhalt-taxonomy.md",11227,"https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Fskills\u002Fce-retune\u002Freferences\u002Fhalt-taxonomy.md",{"path":70,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":71,"sourceHtmlUrl":72},"references\u002Fnoise-floor.md",12142,"https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Fskills\u002Fce-retune\u002Freferences\u002Fnoise-floor.md",{"path":74,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":75,"sourceHtmlUrl":76},"references\u002Fworkflow-shapes.md",11725,"https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Fskills\u002Fce-retune\u002Freferences\u002Fworkflow-shapes.md",[78,85,92,99,106,112],{"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},"01M7D6MM1ZTADSSFC8T6XZQ88C","skill-one","Sample skill","https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fsample-plugin\u002Fskills\u002Fskill-one\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":19},"01M7D6MJTXCGHXFFJF03ZPM5QD","disabled-skill","A skill with model invocation disabled","https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fsample-plugin\u002Fskills\u002Fdisabled-skill\u002FSKILL.md",[],[],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":19},"01M7D6MHMEH8FDQ7NQTD7PHY1Q","claude-only-skill","A skill restricted to Claude Code only","https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fsample-plugin\u002Fskills\u002Fclaude-only-skill\u002FSKILL.md",[],[],{"skillNo":100,"name":101,"nameZh":6,"description":102,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":103,"categories":104,"keywords":105,"publishedAt":19},"01M7D6MG87HVB6HD2TCSFR9MFT","agent-only-skill","An agent-only skill not intended for direct user invocation","https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fsample-plugin\u002Fskills\u002Fagent-only-skill\u002FSKILL.md",[],[],{"skillNo":107,"name":108,"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":109,"categories":110,"keywords":111,"publishedAt":19},"01M7D6MF6TEY1ZYG2E6A580ECS","default-skill","https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fcustom-paths\u002Fskills\u002Fdefault-skill\u002FSKILL.md",[],[],{"skillNo":113,"name":114,"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":115,"categories":116,"keywords":117,"publishedAt":19},"01M7D6MCTYWF4NX7VRRJ07WT8H","custom-skill","https:\u002F\u002Fgithub.com\u002Fcrdant\u002Fcompound-engineering-plugin\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fcustom-paths\u002Fcustom-skills\u002Fcustom-skill\u002FSKILL.md",[],[]]