[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77HMX2JSF7RV5NNB9CM7KDW":3,"skill-same:01M77HMX2JSF7RV5NNB9CM7KDW":49},{"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":27,"commitSha":28,"skillDirectory":29,"forksCount":30,"repositoryHtmlUrl":31,"repositoryHomepageUrl":32,"ownerHtmlUrl":33,"ownerBlogUrl":34,"ownerDescription":6,"updatedAt":35,"install":36,"resources":48},"01M77HMX2JSF7RV5NNB9CM7KDW","steerability-score",null,"One repeatable pass that keeps fak as STEERABLE as it grows — the one scorecard whose every KPI is growth-invariant, so a 2x-larger repo with the same discipline scores the same. Runs the steerability scorecard (tools\u002Fsteerability_scorecard.py) over the working tree, reads the 0-100 steerability index + the advisory drift signals (coupling hubs, p90 sizes, long-function rate, package drift, churn hot spots), drives the index UP and the worst drift axis DOWN by adding REAL modularity (split a cmd dispatch monolith. Use when this named workflow matches the task.","anthony-chaudhary\u002Ffak","anthony-chaudhary","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F18080164?v=4","01M77GSKCAEKV3JS07J40B94DY","Anthony Chaudhary","01M77GSKD26Z7F4FQQD4VA30VX",48,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fsteerability-score\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: steerability-score\ndescription: One repeatable pass that keeps fak as STEERABLE as it grows — the one scorecard whose every KPI is growth-invariant, so a 2x-larger repo with the same discipline scores the same. Runs the steerability scorecard (tools\u002Fsteerability_scorecard.py) over the working tree, reads the 0-100 steerability index + the advisory drift signals (coupling hubs, p90 sizes, long-function rate, package drift, churn hot spots), drives the index UP and the worst drift axis DOWN by adding REAL modularity (split a cmd dispatch monolith. Use when this named workflow matches the task.\ndisable-model-invocation: false\nuser-invocable: true\nallowed-tools: Read, Bash, Write, Edit, Grep, Glob\nargument-hint: \"[--range HEAD~N..HEAD] [--target INDEX]  (no args = full measure + one improvement iteration)\"\n---\n\n# \u002Fsteerability-score — keep steering effort flat as the repo grows\n\n> **What this does.** Every sibling scorecard reports an absolute *count* of\n> defects, which mechanically climbs as the repo grows — a 3×-bigger tree has ~3×\n> the surfaces. None of them answers the question this skill owns: as fak doubles\n> in size, does the **effort to steer, change, and navigate it stay roughly flat**,\n> and if it drifts, do we *know* and can we *correct*? This is that pass. It is an\n> instance of the shared [`scorecard`](..\u002Fscorecard\u002FSKILL.md) doctrine, pointed at\n> the one surface measured in *shape*, not *size*.\n\nThe shape: **run the scorecard → read the index + the worst drift axis → raise the\nindex by adding REAL modularity (never gaming) → re-measure and prove the index rose\n→ commit ONLY your lane by explicit path.**\n\nThe headline is a 0–100 **steerability index** — a weighted mean of growth-invariant\nKPIs — not a debt pile. That is the design: a count would trend just from getting\nbigger, so it can't answer \"is steering effort *flat*.\" The index can. Drive it UP.\n\n---\n\n## The measure (eleven KPIs, four groups, all growth-invariant)\n\n`tools\u002Fsteerability_scorecard.py` folds these into the index + a `steerability_debt`\ninteger (for control-pane membership). **HARD** KPIs emit debt; **SOFT** KPIs score\nthe index but never emit debt — because their only cheap fix would be gaming.\n\n| Group | KPI | HARD\u002FSOFT | what it measures (growth-invariant form) |\n|---|---|---|---|\n| modularity | `file_size_dist` | SOFT | p90 file length vs a fixed reference (percentile — scale-free) |\n| modularity | `func_size_dist` | SOFT | fraction of functions over the soft length line (a rate) |\n| modularity | `god_file_rate` | SOFT | rate of files over the hard ceiling (code_quality owns the COUNT) |\n| modularity | `god_func_rate` | SOFT | rate of functions over the hard ceiling (code_quality owns the COUNT) |\n| coupling | `fan_in_gini` | SOFT | Gini of the internal-import fan-in graph (flat = steerable) |\n| coupling | `hub_share` | SOFT | the single most-depended-on package's share of all packages |\n| coupling | `dispatch_god_file` | **HARD** | a `cmd\u002F*` dispatch file over the hard ceiling |\n| navigability | `package_doc_frac` | SOFT | fraction of packages with a `\u002F\u002F Package x` doc-comment |\n| correction | `ratchet_present` | **HARD** | the control-pane baseline parses + this scorecard is wired |\n| correction | `worst_pkg_drift` | SOFT | worst package's LOC growth vs the pinned baseline |\n| correction | `churn_concentration` | SOFT, HEAD-relative | Gini of recent-commit churn (a hot spot) |\n\n**Why only two HARD KPIs, both ~0 on a disciplined tree.** This scorecard stays\nORTHOGONAL to `code_quality`: god-files and tests are already *that* scorecard's HARD\ndebt, and both fold into the same control-pane total. Re-emitting them here would\ndouble-count the same monolith in the portfolio sum. So steerability **scores**\nsize\u002Fcoupling on the invariant rate (SOFT) and leaves the raw count to `code_quality`.\nThe only things it emits debt for are the two whose cheapest fix is genuinely real\nwork — splitting a dispatch monolith (`dispatch_god_file`) and committing the\ncorrection ratchet (`ratchet_present`). On a healthy tree both are zero, and the\nsignal lives in the **index** and the **drift signals**. That is correct, not a bug.\n\n`churn_concentration` is the one **HEAD-relative** KPI (the `code_quality.ship_integrity`\nprecedent): it reads recent git history, so its number moves as commits land even on a\nbyte-identical tree. Pin `--range HEAD~N..HEAD` for a stable read. It can never anchor\nthe baseline.\n\n---\n\n## Step 1 — Run the scorecard (it builds your work-list)\n\nFrom the repo root:\n\n```bash\npython tools\u002Fsteerability_scorecard.py            # human scorecard (index + per-KPI + drift signals)\npython tools\u002Fsteerability_scorecard.py --json      # machine payload (the loop uses this)\n```\n\nRead `corpus.index`, `corpus.index_by_group`, and `corpus.breakdown` (per-KPI, worst\nfirst). The per-KPI `soft` arrays are the **drift work-list** — where steerability is\nheading even when no hard debt exists yet. **Record the baseline:** write down\n`index = S` and the worst group, because the whole point is to prove the delta.\n\n## Step 2 — Raise the index worst-axis-first, using REAL structural moves only\n\nAttack the lowest group score (`corpus.index_by_group`), then the lowest KPI in it.\nThe genuine moves, by KPI:\n\n- **`dispatch_god_file` (the one HARD coupling defect).** A `cmd\u002F*` file over the\n  ceiling means every new verb fights the same monolith. Split the verb table into\n  per-command files (the [`modularize`](..\u002Fmodularize\u002FSKILL.md) move: behavior-\n  preserving code motion along real seams). This is the highest-value steerability\n  fix — it directly cuts the blast radius of adding a command.\n- **`hub_share` \u002F `fan_in_gini` (coupling, SOFT but the heart of steerability).** A\n  package imported by a large fraction of the tree is a chokepoint every change routes\n  through. The real fix is to **break the dependency**: extract the narrow interface\n  the importers actually need into a small, stable package, or invert the dependency\n  so the hub no longer has to change when its consumers do. NEVER add a façade\n  re-export package to split the hub's name while the real coupling is unchanged —\n  that games the Gini without improving steerability.\n- **`file_size_dist` \u002F `func_size_dist` (modularity, SOFT).** Split a large file along\n  a concern seam; extract a long function into a named helper. Same discipline as\n  `modularize`. A cosmetic split with no real seam games a percentile and is refused.\n- **`package_doc_frac` (navigability, SOFT).** Add a *real* `\u002F\u002F Package x …` header\n  that says what the package is for and when to reach for it — one sentence a reader\n  or agent uses to orient. NEVER `\u002F\u002F Package x provides x.` spam: that games the\n  fraction without aiding navigation (the `godoc` lesson). If a package's purpose is\n  genuinely obvious from its name, leave it and say so.\n- **`ratchet_present` (the one HARD correction defect).** If it fires, the control-pane\n  baseline is missing\u002Fmalformed or this scorecard isn't wired into the fold. Fix it by\n  committing a real baseline (`python tools\u002Fscorecard_control_pane.py --pin`) and the\n  `SCORECARDS` row — the genuine correction affordance, not a `touch`.\n\n`worst_pkg_drift` and `churn_concentration` are *advisory lenses*, not fix targets:\nthey tell you WHERE to look (a ballooning package, a churn hot spot), and the fix is\none of the moves above applied there.\n\n## Step 3 — Validate every structural change (the honesty gate)\n\nAny code motion (splitting a file, extracting a helper, breaking a coupling edge) must\nkeep behavior identical. Native `go test` is OS-blocked on the Windows dev box, so\nvalidate under WSL:\n\n```bash\ngo build .\u002F...                                   # compiles (catches a broken import graph)\ngo vet .\u002F...\nwsl -e bash -lc 'cd \u002Fmnt\u002Fc\u002Fwork\u002Ffak && go test .\u002F\u003Cchanged-pkg>\u002F... -count=1'\n```\n\nA refactor you have not built and tested is not done. Never edit the frozen ABI\n(`internal\u002Fabi`) or add a dependency to move a number.\n\n## Step 4 — Re-measure and PROVE the index rose\n\n```bash\npython tools\u002Fsteerability_scorecard.py --json\npython tools\u002Fsteerability_scorecard.py --compare baseline.json   # the index\u002Fdebt delta + verdict\n```\n\nState the delta plainly: `index S → S' (+k)`, and the group that moved. Pin `--range`\nto the SAME window you measured the baseline with, so the HEAD-relative churn KPI\ndoesn't masquerade as your improvement. Regenerate the committed snapshot:\n\n```bash\npython tools\u002Fsteerability_scorecard.py --markdown --stamp YYYY-MM-DD > docs\u002FSTEERABILITY-SCORECARD.md\n```\n\n(Use the Bash tool's `>` — it preserves UTF-8; a PowerShell `>` re-encodes to UTF-16\nand mangles the `·`\u002F`×`\u002F`—` glyphs.)\n\n## Step 5 — Commit ONLY your lane, by explicit path\n\n```bash\nfak sync reconcile --apply\nfak commit --path tools\u002Fsteerability_scorecard.py --path tools\u002Fsteerability_scorecard_test.py --path docs\u002FSTEERABILITY-SCORECARD.md -F msg\ndos commit-audit HEAD                             # MUST print [diff-witnessed] \u002F verdict OK\nfak sync push\n```\n\nIf your pass did real refactoring in a kernel package, commit that package in its OWN\nlane (the `modularize` discipline), separate from the scorecard lane. Subject honesty:\na refactor → `refactor(\u003Cscope>):`; a scorecard-only measure\u002Fsnapshot → `chore(steer):`\nor `docs(steer):`. End every ship commit with a `(fak steerability)` (or the touched\nleaf's) trailer so the `dos verify` referee binds it. Stay on `main`; never force-push.\n\n---\n\n## The RSI loop\n\nEach pass: measure → raise the lowest group → prove the index rose → commit witnessed.\nNext pass, the *new* lowest group surfaces — the loop walks steerability UP because the\nindex is re-derived from the tree every time and can't be talked past. Because every\nKPI is growth-invariant, **a clean pass holds the index flat even as the repo grows** —\nwhich is the entire goal: the same level of steerability at 2× the size.\n\n## Anti-gaming laws (the index is only as honest as the pass)\n\n1. **Never add a façade \u002F re-export package to split a coupling hub's name.** The real\n   coupling is unchanged; you've gamed the Gini. Break the dependency for real.\n2. **Never do a cosmetic file\u002Ffunction split with no concern seam** to move a\n   percentile. Split along a real boundary or leave it.\n3. **Never spam `\u002F\u002F Package x provides x.`** to move `package_doc_frac` — it's SOFT for\n   exactly this reason. Write a header that actually orients, or leave it.\n4. **Never re-emit god-file\u002Ftest debt here.** `code_quality` owns that count; this\n   scorecard scores the RATE. Double-counting the same monolith is the wiring bug this\n   orthogonality exists to prevent.\n5. **Never let the HEAD-relative `churn_concentration` masquerade as a structural win.**\n   Pin `--range` to the baseline's window before claiming the index moved.\n6. **Re-measure and `dos commit-audit` your commit** before claiming the index rose.\n\n## When to run this\n\n- To **baseline** steerability (first run records the index + the worst group).\n- After a **structural change**: a new package, a new import edge into a hub, a growing\n  `cmd\u002F*` dispatch file, a package that's ballooning.\n- When the project **feels harder to change than its size warrants** — the index tells\n  you which axis (coupling, modularity, navigability) is the culprit.\n- On a `\u002Floop` cadence to keep steering effort flat as the kernel grows between releases.\n\nThe scorecard is read-only; this skill's only writes are your genuine structural fixes,\n`docs\u002FSTEERABILITY-SCORECARD.md`, and the tool itself. It never edits the frozen ABI and\nnever games a SOFT KPI.\n","# \u002Fsteerability-score — keep steering effort flat as the repo grows\n\n> **What this does.** Every sibling scorecard reports an absolute *count* of\n> defects, which mechanically climbs as the repo grows — a 3×-bigger tree has ~3×\n> the surfaces. None of them answers the question this skill owns: as fak doubles\n> in size, does the **effort to steer, change, and navigate it stay roughly flat**,\n> and if it drifts, do we *know* and can we *correct*? This is that pass. It is an\n> instance of the shared [`scorecard`](..\u002Fscorecard\u002FSKILL.md) doctrine, pointed at\n> the one surface measured in *shape*, not *size*.\n\nThe shape: **run the scorecard → read the index + the worst drift axis → raise the\nindex by adding REAL modularity (never gaming) → re-measure and prove the index rose\n→ commit ONLY your lane by explicit path.**\n\nThe headline is a 0–100 **steerability index** — a weighted mean of growth-invariant\nKPIs — not a debt pile. That is the design: a count would trend just from getting\nbigger, so it can't answer \"is steering effort *flat*.\" The index can. Drive it UP.\n\n---\n\n## The measure (eleven KPIs, four groups, all growth-invariant)\n\n`tools\u002Fsteerability_scorecard.py` folds these into the index + a `steerability_debt`\ninteger (for control-pane membership). **HARD** KPIs emit debt; **SOFT** KPIs score\nthe index but never emit debt — because their only cheap fix would be gaming.\n\n| Group | KPI | HARD\u002FSOFT | what it measures (growth-invariant form) |\n|---|---|---|---|\n| modularity | `file_size_dist` | SOFT | p90 file length vs a fixed reference (percentile — scale-free) |\n| modularity | `func_size_dist` | SOFT | fraction of functions over the soft length line (a rate) |\n| modularity | `god_file_rate` | SOFT | rate of files over the hard ceiling (code_quality owns the COUNT) |\n| modularity | `god_func_rate` | SOFT | rate of functions over the hard ceiling (code_quality owns the COUNT) |\n| coupling | `fan_in_gini` | SOFT | Gini of the internal-import fan-in graph (flat = steerable) |\n| coupling | `hub_share` | SOFT | the single most-depended-on package's share of all packages |\n| coupling | `dispatch_god_file` | **HARD** | a `cmd\u002F*` dispatch file over the hard ceiling |\n| navigability | `package_doc_frac` | SOFT | fraction of packages with a `\u002F\u002F Package x` doc-comment |\n| correction | `ratchet_present` | **HARD** | the control-pane baseline parses + this scorecard is wired |\n| correction | `worst_pkg_drift` | SOFT | worst package's LOC growth vs the pinned baseline |\n| correction | `churn_concentration` | SOFT, HEAD-relative | Gini of recent-commit churn (a hot spot) |\n\n**Why only two HARD KPIs, both ~0 on a disciplined tree.** This scorecard stays\nORTHOGONAL to `code_quality`: god-files and tests are already *that* scorecard's HARD\ndebt, and both fold into the same control-pane total. Re-emitting them here would\ndouble-count the same monolith in the portfolio sum. So steerability **scores**\nsize\u002Fcoupling on the invariant rate (SOFT) and leaves the raw count to `code_quality`.\nThe only things it emits debt for are the two whose cheapest fix is genuinely real\nwork — splitting a dispatch monolith (`dispatch_god_file`) and committing the\ncorrection ratchet (`ratchet_present`). On a healthy tree both are zero, and the\nsignal lives in the **index** and the **drift signals**. That is correct, not a bug.\n\n`churn_concentration` is the one **HEAD-relative** KPI (the `code_quality.ship_integrity`\nprecedent): it reads recent git history, so its number moves as commits land even on a\nbyte-identical tree. Pin `--range HEAD~N..HEAD` for a stable read. It can never anchor\nthe baseline.\n\n---\n\n## Step 1 — Run the scorecard (it builds your work-list)\n\nFrom the repo root:\n\n```bash\npython tools\u002Fsteerability_scorecard.py            # human scorecard (index + per-KPI + drift signals)\npython tools\u002Fsteerability_scorecard.py --json      # machine payload (the loop uses this)\n```\n\nRead `corpus.index`, `corpus.index_by_group`, and `corpus.breakdown` (per-KPI, worst\nfirst). The per-KPI `soft` arrays are the **drift work-list** — where steerability is\nheading even when no hard debt exists yet. **Record the baseline:** write down\n`index = S` and the worst group, because the whole point is to prove the delta.\n\n## Step 2 — Raise the index worst-axis-first, using REAL structural moves only\n\nAttack the lowest group score (`corpus.index_by_group`), then the lowest KPI in it.\nThe genuine moves, by KPI:\n\n- **`dispatch_god_file` (the one HARD coupling defect).** A `cmd\u002F*` file over the\n  ceiling means every new verb fights the same monolith. Split the verb table into\n  per-command files (the [`modularize`](..\u002Fmodularize\u002FSKILL.md) move: behavior-\n  preserving code motion along real seams). This is the highest-value steerability\n  fix — it directly cuts the blast radius of adding a command.\n- **`hub_share` \u002F `fan_in_gini` (coupling, SOFT but the heart of steerability).** A\n  package imported by a large fraction of the tree is a chokepoint every change routes\n  through. The real fix is to **break the dependency**: extract the narrow interface\n  the importers actually need into a small, stable package, or invert the dependency\n  so the hub no longer has to change when its consumers do. NEVER add a façade\n  re-export package to split the hub's name while the real coupling is unchanged —\n  that games the Gini without improving steerability.\n- **`file_size_dist` \u002F `func_size_dist` (modularity, SOFT).** Split a large file along\n  a concern seam; extract a long function into a named helper. Same discipline as\n  `modularize`. A cosmetic split with no real seam games a percentile and is refused.\n- **`package_doc_frac` (navigability, SOFT).** Add a *real* `\u002F\u002F Package x …` header\n  that says what the package is for and when to reach for it — one sentence a reader\n  or agent uses to orient. NEVER `\u002F\u002F Package x provides x.` spam: that games the\n  fraction without aiding navigation (the `godoc` lesson). If a package's purpose is\n  genuinely obvious from its name, leave it and say so.\n- **`ratchet_present` (the one HARD correction defect).** If it fires, the control-pane\n  baseline is missing\u002Fmalformed or this scorecard isn't wired into the fold. Fix it by\n  committing a real baseline (`python tools\u002Fscorecard_control_pane.py --pin`) and the\n  `SCORECARDS` row — the genuine correction affordance, not a `touch`.\n\n`worst_pkg_drift` and `churn_concentration` are *advisory lenses*, not fix targets:\nthey tell you WHERE to look (a ballooning package, a churn hot spot), and the fix is\none of the moves above applied there.\n\n## Step 3 — Validate every structural change (the honesty gate)\n\nAny code motion (splitting a file, extracting a helper, breaking a coupling edge) must\nkeep behavior identical. Native `go test` is OS-blocked on the Windows dev box, so\nvalidate under WSL:\n\n```bash\ngo build .\u002F...                                   # compiles (catches a broken import graph)\ngo vet .\u002F...\nwsl -e bash -lc 'cd \u002Fmnt\u002Fc\u002Fwork\u002Ffak && go test .\u002F\u003Cchanged-pkg>\u002F... -count=1'\n```\n\nA refactor you have not built and tested is not done. Never edit the frozen ABI\n(`internal\u002Fabi`) or add a dependency to move a number.\n\n## Step 4 — Re-measure and PROVE the index rose\n\n```bash\npython tools\u002Fsteerability_scorecard.py --json\npython tools\u002Fsteerability_scorecard.py --compare baseline.json   # the index\u002Fdebt delta + verdict\n```\n\nState the delta plainly: `index S → S' (+k)`, and the group that moved. Pin `--range`\nto the SAME window you measured the baseline with, so the HEAD-relative churn KPI\ndoesn't masquerade as your improvement. Regenerate the committed snapshot:\n\n```bash\npython tools\u002Fsteerability_scorecard.py --markdown --stamp YYYY-MM-DD > docs\u002FSTEERABILITY-SCORECARD.md\n```\n\n(Use the Bash tool's `>` — it preserves UTF-8; a PowerShell `>` re-encodes to UTF-16\nand mangles the `·`\u002F`×`\u002F`—` glyphs.)\n\n## Step 5 — Commit ONLY your lane, by explicit path\n\n```bash\nfak sync reconcile --apply\nfak commit --path tools\u002Fsteerability_scorecard.py --path tools\u002Fsteerability_scorecard_test.py --path docs\u002FSTEERABILITY-SCORECARD.md -F msg\ndos commit-audit HEAD                             # MUST print [diff-witnessed] \u002F verdict OK\nfak sync push\n```\n\nIf your pass did real refactoring in a kernel package, commit that package in its OWN\nlane (the `modularize` discipline), separate from the scorecard lane. Subject honesty:\na refactor → `refactor(\u003Cscope>):`; a scorecard-only measure\u002Fsnapshot → `chore(steer):`\nor `docs(steer):`. End every ship commit with a `(fak steerability)` (or the touched\nleaf's) trailer so the `dos verify` referee binds it. Stay on `main`; never force-push.\n\n---\n\n## The RSI loop\n\nEach pass: measure → raise the lowest group → prove the index rose → commit witnessed.\nNext pass, the *new* lowest group surfaces — the loop walks steerability UP because the\nindex is re-derived from the tree every time and can't be talked past. Because every\nKPI is growth-invariant, **a clean pass holds the index flat even as the repo grows** —\nwhich is the entire goal: the same level of steerability at 2× the size.\n\n## Anti-gaming laws (the index is only as honest as the pass)\n\n1. **Never add a façade \u002F re-export package to split a coupling hub's name.** The real\n   coupling is unchanged; you've gamed the Gini. Break the dependency for real.\n2. **Never do a cosmetic file\u002Ffunction split with no concern seam** to move a\n   percentile. Split along a real boundary or leave it.\n3. **Never spam `\u002F\u002F Package x provides x.`** to move `package_doc_frac` — it's SOFT for\n   exactly this reason. Write a header that actually orients, or leave it.\n4. **Never re-emit god-file\u002Ftest debt here.** `code_quality` owns that count; this\n   scorecard scores the RATE. Double-counting the same monolith is the wiring bug this\n   orthogonality exists to prevent.\n5. **Never let the HEAD-relative `churn_concentration` masquerade as a structural win.**\n   Pin `--range` to the baseline's window before claiming the index moved.\n6. **Re-measure and `dos commit-audit` your commit** before claiming the index rose.\n\n## When to run this\n\n- To **baseline** steerability (first run records the index + the worst group).\n- After a **structural change**: a new package, a new import edge into a hub, a growing\n  `cmd\u002F*` dispatch file, a package that's ballooning.\n- When the project **feels harder to change than its size warrants** — the index tells\n  you which axis (coupling, modularity, navigability) is the culprit.\n- On a `\u002Floop` cadence to keep steering effort flat as the kernel grows between releases.\n\nThe scorecard is read-only; this skill's only writes are your genuine structural fixes,\n`docs\u002FSTEERABILITY-SCORECARD.md`, and the tool itself. It never edits the frozen ABI and\nnever games a SOFT KPI.\n",{"name":5,"description":7,"allowed-tools":24,"argument-hint":25,"user-invocable":26,"disable-model-invocation":20},"Read, Bash, Write, Edit, Grep, Glob","[--range HEAD~N..HEAD] [--target INDEX]  (no args = full measure + one improvement iteration)",true,"main","f565b7b7db63e79649f3a02c08108132a86c5d05",".claude\u002Fskills\u002Fsteerability-score",16,"https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak","https:\u002F\u002Fanthony-chaudhary.github.io\u002Ffak\u002Fshowcase.html","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary","https:\u002F\u002Flinkedin.com\u002Fin\u002Fanthonychaudhary","2026-11-15 03:06:00",{"source":8,"skillName":5,"commands":37,"prompt":47,"zipAvailable":26},[38,41,44],{"tool":39,"command":40},"npx","npx skills add anthony-chaudhary\u002Ffak --skill steerability-score",{"tool":42,"command":43},"bunx","bunx skills add anthony-chaudhary\u002Ffak --skill steerability-score",{"tool":45,"command":46},"pnpm","pnpm dlx skills add anthony-chaudhary\u002Ffak --skill steerability-score","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fsteerability-score\u002FSKILL.md",[],[50,57,66,73,80,85],{"skillNo":51,"name":52,"nameZh":6,"description":53,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":54,"categories":55,"keywords":56,"publishedAt":19},"01M78V4059KRP2T88YR6Z05MJQ","serve-receipt","On-demand research protocol + deterministic receipt for serving a Hugging Face model on a concrete GPU target — fan out research lanes (model facts\u002Ffiles, engine support matrix, serving recipes, KV\u002Fstate memory math), compute the tiered feasibility verdict with `fak serve-receipt` (never narrated math), and emit a witnessed launch receipt. Use when asked for an \"initial config\", \"serving receipt\", or \"how do I serve \u003Cmodel> on \u003CGPUs>\".","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fserve-receipt\u002FSKILL.md",[],[],{"skillNo":58,"name":59,"nameZh":6,"description":60,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":61,"licenseFileUrl":62,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":19},"01M77HRGKQV2CJ2K6ZFN4YQ32K","caveman","Ultra-compressed communication mode. Cuts output tokens 65% (measured) by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says \"caveman mode\", \"talk like caveman\", \"use caveman\", \"less tokens\", \"be brief\", or invokes \u002Fcaveman. Also auto-triggers when token efficiency is requested.","MIT","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002Finternal\u002Fsyspromptmmu\u002Fthird_party\u002Fcaveman\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002Finternal\u002Fsyspromptmmu\u002Fthird_party\u002Fcaveman\u002FSKILL.md",[],[],{"skillNo":67,"name":68,"nameZh":6,"description":69,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":19},"01M77HRAY3Y3BE7APN5PVJF7NV","quality-report","Score the repo quality and emit a report card.","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002Finternal\u002Fctxplans\u002Ftestdata\u002Frepo\u002F.claude\u002Fskills\u002Fquality-report\u002FSKILL.md",[],[],{"skillNo":74,"name":75,"nameZh":6,"description":76,"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},"01M77HR9807KHAX13DG9PW2K2Q","ctx-overlay","Compact the context window and prune the memory store when a session grows too long.","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002Finternal\u002Fctxplans\u002Ftestdata\u002Frepo\u002F.claude\u002Fskills\u002Fctx-overlay\u002FSKILL.md",[],[],{"skillNo":81,"name":68,"nameZh":6,"description":69,"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},"01M77HR7XQ9X8F2CY9Z4F2ZEZ3","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002Finternal\u002Fctxplanlint\u002Ftestdata\u002Frepo\u002F.claude\u002Fskills\u002Fquality-report\u002FSKILL.md",[],[],{"skillNo":86,"name":75,"nameZh":6,"description":76,"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},"01M77HR5BVKCTR1WNJGRBSDNX7","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002Finternal\u002Fctxplanlint\u002Ftestdata\u002Frepo\u002F.claude\u002Fskills\u002Fctx-overlay\u002FSKILL.md",[],[]]