[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77KP1T1WD17BH5FEGWSVVSE":3,"skill-same:01M77KP1T1WD17BH5FEGWSVVSE":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":29,"commitSha":30,"skillDirectory":31,"forksCount":32,"repositoryHtmlUrl":33,"repositoryHomepageUrl":6,"ownerHtmlUrl":34,"ownerBlogUrl":35,"ownerDescription":36,"updatedAt":37,"install":38,"resources":50},"01M77KP1T1WD17BH5FEGWSVVSE","methodology",null,"When the bundled claude-api skill resolves in this session, prefer its hillclimb for sweeping model and effort over a measurable suite; this skill before that suite or its criteria exist. Answers LLM-evaluation design questions from Anthropic's evaluation guidance: success criteria (including turning a vague quality goal into a measurable target before any data exists), eval-suite design, grading methods, and the eval route for each part of a repository. Use when: an eval success criterion, metric or target is being written or rewritten, with or without data (load this before answering), 'how do I eval this', 'LLM eval', 'measure prompt quality', 'LLM judge', 'model-graded eval', 'golden answer', 'grading rubric', 'eval grading method', 'exact match vs LLM-graded', 'how many eval cases', 'evals for a service that calls the Messages API', 'evals for the whole repo'. Knowledge, not a runner; scaffolding a suite is \u002Fevals:design, and running a plugin's suite against a no-plugin baseline is \u002Fevals:plugin-eval.","melodic-software\u002Fclaude-code-plugins","melodic-software","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F58273638?v=4","01M77JVQTXNGKCE4MDFY2BC936","Melodic Software","01M77JVQVN9QDWKAH6W03ANP76",30,"MIT","https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fevals\u002Fskills\u002Fmethodology\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\ndescription: \"When the bundled claude-api skill resolves in this session, prefer its hillclimb for sweeping model and effort over a measurable suite; this skill before that suite or its criteria exist. Answers LLM-evaluation design questions from Anthropic's evaluation guidance: success criteria (including turning a vague quality goal into a measurable target before any data exists), eval-suite design, grading methods, and the eval route for each part of a repository. Use when: an eval success criterion, metric or target is being written or rewritten, with or without data (load this before answering), 'how do I eval this', 'LLM eval', 'measure prompt quality', 'LLM judge', 'model-graded eval', 'golden answer', 'grading rubric', 'eval grading method', 'exact match vs LLM-graded', 'how many eval cases', 'evals for a service that calls the Messages API', 'evals for the whole repo'. Knowledge, not a runner; scaffolding a suite is \u002Fevals:design, and running a plugin's suite against a no-plugin baseline is \u002Fevals:plugin-eval.\"\nargument-hint: \"[question or concept]\"\nuser-invocable: true\ndisable-model-invocation: false\nmetadata:\n  workflow-stage: test\n  summary: Answer LLM-evaluation design questions from Anthropic's official guidance\n---\n\n# LLM evaluation methodology\n\nThis page states the rules; answer from it. It runs, scores, scaffolds and edits nothing.\n\n- **Pointer**: when a specific must be current, fetch\n  [Define your success criteria](https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Ftest-and-evaluate\u002Fdevelop-tests#define-your-success-criteria)\n  and [Build evaluations](https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Ftest-and-evaluate\u002Fdevelop-tests#build-evaluations)\n  live; the evals cookbook is `anthropics\u002Fclaude-cookbooks` `misc\u002Fbuilding_evals.ipynb`.\n- **As of**: 2026-08-08 (distilled); pointer anchors verified 2026-10-11\n- **Recheck trigger**: either section is retitled, moves, or changes the guidance these rules rest\n  on.\n\n## Background files\n\nEach is background for a human reader, with its own source stamp. For runnable code, or a\nspecific that must be current, check the source page instead.\n\n| Topic | File |\n|---|---|\n| Success criteria: the four properties, quantifying hazy qualities, metric menu, dimensions | [success-criteria.md](reference\u002Fsuccess-criteria.md) |\n| Eval anatomy, golden answers, edge-case taxonomy, volume over polish, generating cases, effort and model sweeps | [eval-design.md](reference\u002Feval-design.md) |\n| Grading ladder, LLM-grader rubrics, constrained verdicts, testing the grader first | [grading.md](reference\u002Fgrading.md) |\n| Recipes: exact match, cosine similarity, ROUGE-L, Likert, binary, ordinal | [recipes.md](reference\u002Frecipes.md) |\n| Improving against an existing suite with the bundled `hillclimb`: who starts it, the step map for a plugin suite, cost as the goal | [hillclimb.md](reference\u002Fhillclimb.md) |\n| This repository's defaults where sources disagree: train\u002Ftest split, interval method, repeat count, rubric form, and their settings | [local-decisions.md](reference\u002Flocal-decisions.md) |\n\n## Quick decision guide\n\n- \"Where do I start?\" For a Claude API app, with `\u002Fclaude-api build-eval` (see\n  [Route by repository kind](#route-by-repository-kind)). Otherwise, define measurable success\n  criteria first; evals test against them; only then iterate on prompts.\n- \"Is this criterion good?\" It names a specific quality, a number or defined scale, a realistic\n  target, and ties to a user need. The target number is realistic only when a measured baseline, a\n  prior result, a benchmark, or expert review justifies it; how severe a miss would be, or that a\n  grader can check the number, does not. \"Good performance\" fails all four.\n- \"Rewrite this criterion\" Answer with this template, every slot filled:\n\n  ```text\n  quality: \u003Cthe specific behavior measured, and the user need it serves>\n  measure: \u003Cthe metric: a rate, a count or a defined scale>\n  trials: \u003Cthe trial set and its size, for example \"500 conversations\">\n  target: \u003Cone threshold, taken from the basis>\n  basis: \u003Cmeasured baseline | prior result | benchmark | expert review | baseline-relative>\n  ```\n\n  Each threshold gets its own block: a second bar on the same quality, such as a stricter one for\n  its worst failures, is a second block with its own `basis:`. The `basis:` slot takes one of\n  those five values, and the target must follow from it. The first four count only when the user\n  supplied the figure or agreement; name it. When none applies, the basis is `baseline-relative`\n  and the target is a fraction of the current rate, written without inventing its value (for\n  example \"at most half the current rate over 500 trials\"). An absolute number, zero included,\n  has no place in a `baseline-relative` block, so a \"never\" bar is written relative too (\"at most\n  a tenth of the current rate of refunds sent to the wrong account\"). Never write a target as a\n  placeholder (X\u002FY\u002FZ, \"set the target later\") or a bare direction (\"lower than the baseline\").\n  Severity, or that a grader can check the number, is not a basis. Never present a made-up\n  baseline figure or expert agreement as fact.\n- \"Which grading method?\" The fastest, most reliable, most scalable that fits: code-based if the\n  output can be constrained to allow it; LLM-graded for judgment; human only as a last resort.\n- \"Can I automate this seemingly subjective eval?\" Usually. Constrain the output format,\n  reformat to multiple choice, or use an LLM grader with a tight rubric and constrained verdict.\n- \"How many cases?\" Prefer volume with automated grading over a few hand-graded showpieces;\n  generate more from a baseline set with Claude, human-reviewed. Volume means cheaper grading,\n  never easier cases.\n- \"Can I trust my LLM grader?\" Only after reading samples of its verdicts against your own\n  judgment; and grade with a different model than the one that generated the output.\n- \"How do I grade an agent?\" On the environment state it left, with a grader it cannot reach;\n  see [Agent evals: grader hygiene](#agent-evals-grader-hygiene).\n- \"One metric or several?\" Several. Most use cases need multidimensional criteria (fidelity,\n  safety, latency, cost); a single headline metric hides regressions.\n\nThe target-number and rewrite rules are this repository's reading of the Achievable property.\n\n- **Pointer**: for what grounds a target, see\n  \u003Chttps:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Ftest-and-evaluate\u002Fdevelop-tests#define-your-success-criteria>.\n\n## Agent evals: grader hygiene\n\nWhen the thing under test is an agent that acts over many turns and changes an environment:\n\n- Grade the outcome, the state the agent left the environment in, not the transcript's claim that\n  the work is done. An agent saying the record exists is not evidence that it does; the grader\n  checks the record.\n- Keep graders out of the agent's reach. A check the agent can read, edit or special-case is one\n  it can pass without doing the task, so graders and their expected values sit where the agent\n  under test cannot change them.\n- Read transcripts anyway, in samples, to catch an agent gaming a grader that passed it.\n\n- **Pointer**: Anthropic's \"Demystifying evals for AI agents\" (2026-01-09)\n  \u003Chttps:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Fdemystifying-evals-for-ai-agents>. Read through a fetch\n  summary with quoted sentences, not a raw-text check.\n\n## Route by repository kind\n\nDecide the kind from the repository, then name the route. The user starts `build-eval` and\n`hillclimb`; this skill never does.\n\n| Repository | Route |\n|---|---|\n| A Claude API app (its code calls Claude through the Anthropic SDK or API) | With no eval yet, the first step is `\u002Fclaude-api build-eval`, which the user types. Lead the answer with it; do not lay out criteria, cases and graders by hand before it. With an eval in place, `\u002Fclaude-api hillclimb` improves against it |\n| A skill or plugin repository (a `.claude-plugin\u002Fplugin.json` or `SKILL.md` files) | `claude plugin eval`, through `\u002Fevals:design` to scaffold the suite and `\u002Fevals:plugin-eval` to run it |\n| Both | Report both routes, each for its own part of the repository |\n| No model in the loop | Say that LLM eval design does not apply, and stop |\n\n## Boundary\n\n- `\u002Fevals:design` interviews for criteria and scaffolds a suite; `\u002Fevals:plugin-eval` runs a\n  plugin's suite against a no-plugin baseline and reads the delta; `\u002Fevals:validate` checks a\n  suite's case files before a run.\n- The bundled `claude-api` skill's `hillclimb` searches model, effort and prompt against a suite\n  you already have. When that skill resolves in this session, prefer its `hillclimb` for the\n  search itself; prefer this skill when the suite does not exist yet or its criteria are not yet\n  measurable. The two chain: design here, then hand the\n  suite to the search. `hillclimb` applies accepted configuration and prompt changes to the\n  application under test, so never chain into a `hillclimb` run on this skill's behalf; name the\n  option and let the user invoke it.\n- Availability is never assumed: bundled surfaces are gated by settings, environment, plan, and\n  host, so state what to do when the surface resolves, never that it is present.\n\n- **Pointer**: for the subcommands, see\n  \u003Chttps:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fskills#work-on-claude-api-projects>; for their guides, see\n  [`eval-hillclimb.md`](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fskills\u002Fblob\u002F8a1541c4a3ffa5a20a5a91de0dcf3f0bab1d1ef4\u002Fskills\u002Fclaude-api\u002Fshared\u002Fevals\u002Feval-hillclimb.md)\n  and\n  [`build-eval.md`](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fskills\u002Fblob\u002F8a1541c4a3ffa5a20a5a91de0dcf3f0bab1d1ef4\u002Fskills\u002Fclaude-api\u002Fshared\u002Fevals\u002Fbuild-eval.md)\n  at the pinned commit.\n- **As of**: 2026-10-01\n- **Recheck trigger**: the Claude API skill docs page\n  (\u003Chttps:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fagents-and-tools\u002Fagent-skills\u002Fclaude-api-skill>) lists\n  `build-eval` and `hillclimb`; then repoint there.\n\n## Maintainer `update` action\n\n`\u002Fevals:methodology update` is a maintainer-only drift check. Re-fetch the source page (raw\nmarkdown) and the cookbook notebook, diff against the four distilled files (success-criteria,\neval-design, grading, recipes), apply content corrections, and refresh every \"fetched YYYY-MM-DD\"\nstamp. `local-decisions.md` and `hillclimb.md` are this repository's own records, outside the\naction, with their own recheck triggers. Keep the one pointer line to `local-decisions.md` in each\ndistilled file through a refresh.\n\n## Next\n\n- Criteria settled and a suite is to be written: `\u002Fevals:design \u003Ctarget>`.\n- A plugin suite exists and the question is what it measures: `\u002Fevals:plugin-eval run \u003Ctarget>`.\n\n## Gotchas\n\n- The reference files are a distillation with fetch-date stamps, not the source: for runnable\n  recipe code or any detail that must be current, re-fetch the source page. Its code samples and\n  model names move with releases.\n- Do not \"verify\" a claim about the guidance against this skill's own spokes; the spokes are the\n  derived copy. Verification means fetching the upstream page.\n","# LLM evaluation methodology\n\nThis page states the rules; answer from it. It runs, scores, scaffolds and edits nothing.\n\n- **Pointer**: when a specific must be current, fetch\n  [Define your success criteria](https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Ftest-and-evaluate\u002Fdevelop-tests#define-your-success-criteria)\n  and [Build evaluations](https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Ftest-and-evaluate\u002Fdevelop-tests#build-evaluations)\n  live; the evals cookbook is `anthropics\u002Fclaude-cookbooks` `misc\u002Fbuilding_evals.ipynb`.\n- **As of**: 2026-08-08 (distilled); pointer anchors verified 2026-10-11\n- **Recheck trigger**: either section is retitled, moves, or changes the guidance these rules rest\n  on.\n\n## Background files\n\nEach is background for a human reader, with its own source stamp. For runnable code, or a\nspecific that must be current, check the source page instead.\n\n| Topic | File |\n|---|---|\n| Success criteria: the four properties, quantifying hazy qualities, metric menu, dimensions | [success-criteria.md](reference\u002Fsuccess-criteria.md) |\n| Eval anatomy, golden answers, edge-case taxonomy, volume over polish, generating cases, effort and model sweeps | [eval-design.md](reference\u002Feval-design.md) |\n| Grading ladder, LLM-grader rubrics, constrained verdicts, testing the grader first | [grading.md](reference\u002Fgrading.md) |\n| Recipes: exact match, cosine similarity, ROUGE-L, Likert, binary, ordinal | [recipes.md](reference\u002Frecipes.md) |\n| Improving against an existing suite with the bundled `hillclimb`: who starts it, the step map for a plugin suite, cost as the goal | [hillclimb.md](reference\u002Fhillclimb.md) |\n| This repository's defaults where sources disagree: train\u002Ftest split, interval method, repeat count, rubric form, and their settings | [local-decisions.md](reference\u002Flocal-decisions.md) |\n\n## Quick decision guide\n\n- \"Where do I start?\" For a Claude API app, with `\u002Fclaude-api build-eval` (see\n  [Route by repository kind](#route-by-repository-kind)). Otherwise, define measurable success\n  criteria first; evals test against them; only then iterate on prompts.\n- \"Is this criterion good?\" It names a specific quality, a number or defined scale, a realistic\n  target, and ties to a user need. The target number is realistic only when a measured baseline, a\n  prior result, a benchmark, or expert review justifies it; how severe a miss would be, or that a\n  grader can check the number, does not. \"Good performance\" fails all four.\n- \"Rewrite this criterion\" Answer with this template, every slot filled:\n\n  ```text\n  quality: \u003Cthe specific behavior measured, and the user need it serves>\n  measure: \u003Cthe metric: a rate, a count or a defined scale>\n  trials: \u003Cthe trial set and its size, for example \"500 conversations\">\n  target: \u003Cone threshold, taken from the basis>\n  basis: \u003Cmeasured baseline | prior result | benchmark | expert review | baseline-relative>\n  ```\n\n  Each threshold gets its own block: a second bar on the same quality, such as a stricter one for\n  its worst failures, is a second block with its own `basis:`. The `basis:` slot takes one of\n  those five values, and the target must follow from it. The first four count only when the user\n  supplied the figure or agreement; name it. When none applies, the basis is `baseline-relative`\n  and the target is a fraction of the current rate, written without inventing its value (for\n  example \"at most half the current rate over 500 trials\"). An absolute number, zero included,\n  has no place in a `baseline-relative` block, so a \"never\" bar is written relative too (\"at most\n  a tenth of the current rate of refunds sent to the wrong account\"). Never write a target as a\n  placeholder (X\u002FY\u002FZ, \"set the target later\") or a bare direction (\"lower than the baseline\").\n  Severity, or that a grader can check the number, is not a basis. Never present a made-up\n  baseline figure or expert agreement as fact.\n- \"Which grading method?\" The fastest, most reliable, most scalable that fits: code-based if the\n  output can be constrained to allow it; LLM-graded for judgment; human only as a last resort.\n- \"Can I automate this seemingly subjective eval?\" Usually. Constrain the output format,\n  reformat to multiple choice, or use an LLM grader with a tight rubric and constrained verdict.\n- \"How many cases?\" Prefer volume with automated grading over a few hand-graded showpieces;\n  generate more from a baseline set with Claude, human-reviewed. Volume means cheaper grading,\n  never easier cases.\n- \"Can I trust my LLM grader?\" Only after reading samples of its verdicts against your own\n  judgment; and grade with a different model than the one that generated the output.\n- \"How do I grade an agent?\" On the environment state it left, with a grader it cannot reach;\n  see [Agent evals: grader hygiene](#agent-evals-grader-hygiene).\n- \"One metric or several?\" Several. Most use cases need multidimensional criteria (fidelity,\n  safety, latency, cost); a single headline metric hides regressions.\n\nThe target-number and rewrite rules are this repository's reading of the Achievable property.\n\n- **Pointer**: for what grounds a target, see\n  \u003Chttps:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Ftest-and-evaluate\u002Fdevelop-tests#define-your-success-criteria>.\n\n## Agent evals: grader hygiene\n\nWhen the thing under test is an agent that acts over many turns and changes an environment:\n\n- Grade the outcome, the state the agent left the environment in, not the transcript's claim that\n  the work is done. An agent saying the record exists is not evidence that it does; the grader\n  checks the record.\n- Keep graders out of the agent's reach. A check the agent can read, edit or special-case is one\n  it can pass without doing the task, so graders and their expected values sit where the agent\n  under test cannot change them.\n- Read transcripts anyway, in samples, to catch an agent gaming a grader that passed it.\n\n- **Pointer**: Anthropic's \"Demystifying evals for AI agents\" (2026-01-09)\n  \u003Chttps:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Fdemystifying-evals-for-ai-agents>. Read through a fetch\n  summary with quoted sentences, not a raw-text check.\n\n## Route by repository kind\n\nDecide the kind from the repository, then name the route. The user starts `build-eval` and\n`hillclimb`; this skill never does.\n\n| Repository | Route |\n|---|---|\n| A Claude API app (its code calls Claude through the Anthropic SDK or API) | With no eval yet, the first step is `\u002Fclaude-api build-eval`, which the user types. Lead the answer with it; do not lay out criteria, cases and graders by hand before it. With an eval in place, `\u002Fclaude-api hillclimb` improves against it |\n| A skill or plugin repository (a `.claude-plugin\u002Fplugin.json` or `SKILL.md` files) | `claude plugin eval`, through `\u002Fevals:design` to scaffold the suite and `\u002Fevals:plugin-eval` to run it |\n| Both | Report both routes, each for its own part of the repository |\n| No model in the loop | Say that LLM eval design does not apply, and stop |\n\n## Boundary\n\n- `\u002Fevals:design` interviews for criteria and scaffolds a suite; `\u002Fevals:plugin-eval` runs a\n  plugin's suite against a no-plugin baseline and reads the delta; `\u002Fevals:validate` checks a\n  suite's case files before a run.\n- The bundled `claude-api` skill's `hillclimb` searches model, effort and prompt against a suite\n  you already have. When that skill resolves in this session, prefer its `hillclimb` for the\n  search itself; prefer this skill when the suite does not exist yet or its criteria are not yet\n  measurable. The two chain: design here, then hand the\n  suite to the search. `hillclimb` applies accepted configuration and prompt changes to the\n  application under test, so never chain into a `hillclimb` run on this skill's behalf; name the\n  option and let the user invoke it.\n- Availability is never assumed: bundled surfaces are gated by settings, environment, plan, and\n  host, so state what to do when the surface resolves, never that it is present.\n\n- **Pointer**: for the subcommands, see\n  \u003Chttps:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fskills#work-on-claude-api-projects>; for their guides, see\n  [`eval-hillclimb.md`](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fskills\u002Fblob\u002F8a1541c4a3ffa5a20a5a91de0dcf3f0bab1d1ef4\u002Fskills\u002Fclaude-api\u002Fshared\u002Fevals\u002Feval-hillclimb.md)\n  and\n  [`build-eval.md`](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fskills\u002Fblob\u002F8a1541c4a3ffa5a20a5a91de0dcf3f0bab1d1ef4\u002Fskills\u002Fclaude-api\u002Fshared\u002Fevals\u002Fbuild-eval.md)\n  at the pinned commit.\n- **As of**: 2026-10-01\n- **Recheck trigger**: the Claude API skill docs page\n  (\u003Chttps:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fagents-and-tools\u002Fagent-skills\u002Fclaude-api-skill>) lists\n  `build-eval` and `hillclimb`; then repoint there.\n\n## Maintainer `update` action\n\n`\u002Fevals:methodology update` is a maintainer-only drift check. Re-fetch the source page (raw\nmarkdown) and the cookbook notebook, diff against the four distilled files (success-criteria,\neval-design, grading, recipes), apply content corrections, and refresh every \"fetched YYYY-MM-DD\"\nstamp. `local-decisions.md` and `hillclimb.md` are this repository's own records, outside the\naction, with their own recheck triggers. Keep the one pointer line to `local-decisions.md` in each\ndistilled file through a refresh.\n\n## Next\n\n- Criteria settled and a suite is to be written: `\u002Fevals:design \u003Ctarget>`.\n- A plugin suite exists and the question is what it measures: `\u002Fevals:plugin-eval run \u003Ctarget>`.\n\n## Gotchas\n\n- The reference files are a distillation with fetch-date stamps, not the source: for runnable\n  recipe code or any detail that must be current, re-fetch the source page. Its code samples and\n  model names move with releases.\n- Do not \"verify\" a claim about the guidance against this skill's own spokes; the spokes are the\n  derived copy. Verification means fetching the upstream page.\n",{"metadata":24,"description":7,"argument-hint":27,"user-invocable":28,"disable-model-invocation":20},{"summary":25,"workflow-stage":26},"Answer LLM-evaluation design questions from Anthropic's official guidance","test","[question or concept]",true,"main","d0e68c3f36b92a1d6381f0ba0060d7a0b8b886c5","plugins\u002Fevals\u002Fskills\u002Fmethodology",7,"https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins","https:\u002F\u002Fgithub.com\u002Fmelodic-software","https:\u002F\u002Fmelodicsoftware.com","Independent software studio building developer tools, automation, and application services, with roots in music.","2026-11-15 05:15:06",{"source":8,"skillName":5,"commands":39,"prompt":49,"zipAvailable":28},[40,43,46],{"tool":41,"command":42},"npx","npx skills add melodic-software\u002Fclaude-code-plugins --skill methodology",{"tool":44,"command":45},"bunx","bunx skills add melodic-software\u002Fclaude-code-plugins --skill methodology",{"tool":47,"command":48},"pnpm","pnpm dlx skills add melodic-software\u002Fclaude-code-plugins --skill methodology","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fevals\u002Fskills\u002Fmethodology\u002FSKILL.md",[51,57,61,65,69,73],{"path":52,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":55,"sourceHtmlUrl":56},"reference\u002Feval-design.md","REFERENCE","参考文档",5462,"https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fevals\u002Fskills\u002Fmethodology\u002Freference\u002Feval-design.md",{"path":58,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":59,"sourceHtmlUrl":60},"reference\u002Fgrading.md",4132,"https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fevals\u002Fskills\u002Fmethodology\u002Freference\u002Fgrading.md",{"path":62,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":63,"sourceHtmlUrl":64},"reference\u002Fhillclimb.md",9134,"https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fevals\u002Fskills\u002Fmethodology\u002Freference\u002Fhillclimb.md",{"path":66,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":67,"sourceHtmlUrl":68},"reference\u002Flocal-decisions.md",6658,"https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fevals\u002Fskills\u002Fmethodology\u002Freference\u002Flocal-decisions.md",{"path":70,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":71,"sourceHtmlUrl":72},"reference\u002Frecipes.md",3600,"https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fevals\u002Fskills\u002Fmethodology\u002Freference\u002Frecipes.md",{"path":74,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":75,"sourceHtmlUrl":76},"reference\u002Fsuccess-criteria.md",4303,"https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fevals\u002Fskills\u002Fmethodology\u002Freference\u002Fsuccess-criteria.md",[78,85,92,99,104,111],{"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},"01M7DYYDG31B9YX7N9MB9QW2GW","track","Backlog reads through the bound tracker: search open and closed items, view one item with its comments and state history. Use when another skill searches or reads tracker items.","https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fsession-flow\u002Fevals\u002Frecall-still-reports-reopened-bug\u002Fwork-items-stub\u002Fskills\u002Ftrack\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},"01M7DYXVBAVG3NV30PA7GKW9Q6","reviewer-media","Reviewer media for one pull request: decides whether it needs a narrated walkthrough, builds the storyboard from \u002Freview:explain-change's digest, has \u002Fexplainer-video:produce render it and hands the manifest to \u002Fsource-control:pull-request media. Use when asked for 'reviewer media', 'a walkthrough video for this PR', 'a produced video of this change', or when pull-request ready or monitor hands a PR over.","https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Freview\u002Fskills\u002Freviewer-media\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},"01M7DYRZ5YD44AMYH26YR25GZ8","route","Model and effort per agent role (orchestrator, worker, verifier, retrieval) for a multi-agent run, resolved through the bundled defaults and the user, team and personal override config layers into fenced JSON with single-agent and fan-out variants and the layer behind each value. Use when asked 'which model for the subagents', 'route the roles' or 'model routing', or before a skill passes a role's model on a spawn.","https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fdiscovery\u002Fevals\u002Fex-model-pin-without-override\u002Fmulti-agent-stub\u002Fskills\u002Froute\u002FSKILL.md",[],[],{"skillNo":100,"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":101,"categories":102,"keywords":103,"publishedAt":19},"01M7DYRTVT4FWTGDJY0HCCFJMZ","https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fdiscovery\u002Fevals\u002Fex-model-override-from-route\u002Fmulti-agent-stub\u002Fskills\u002Froute\u002FSKILL.md",[],[],{"skillNo":105,"name":106,"nameZh":6,"description":107,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":108,"categories":109,"keywords":110,"publishedAt":19},"01M7DYNZC8PNW3T27V2ADKECKJ","principles","Module-design vocabulary for architecture findings, and the checks built on it. Use when: 'what is a deep module', 'what is a seam', 'deep vs shallow module', 'what is the deletion test', 'leverage and locality', 'module design vocabulary', 'which word do we use for X in architecture findings'. Scanning a codebase for friction is \u002Farchitecture:improve.","https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Farchitecture\u002Fskills\u002Fprinciples\u002FSKILL.md",[],[],{"skillNo":112,"name":113,"nameZh":6,"description":114,"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},"01M7DYNMCZNEWBFHV9KTSS6H60","write","Structured five-field bug report (title, steps to reproduce, expected vs actual, justified severity, suggested fix location) from an informal description, read-only. Use when the user names a defect they observed, or asks for one written up ('report a bug', 'file a bug', 'write this up as a bug').","https:\u002F\u002Fgithub.com\u002Fmelodic-software\u002Fclaude-code-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Farchitecture\u002Fevals\u002Fdeepening-defect-goes-to-bugs-write\u002Fbugs-stub\u002Fskills\u002Fwrite\u002FSKILL.md",[],[]]