[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7BNK92FNP682W39436HRPR0":3,"skill-same:01M7BNK92FNP682W39436HRPR0":70},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"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":24,"commitSha":25,"skillDirectory":26,"forksCount":13,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":29,"ownerDescription":30,"updatedAt":31,"install":32,"resources":45},"01M7BNK92FNP682W39436HRPR0","spec",null,"Create or modify spec files following project methodology, and drive SDD (spec-first development). Use when writing, editing, or reviewing specs\u002F files; before implementing any feature or behavior change that affects a contract documented in specs\u002F; and whenever the user asks anything about spec-first development — the SDD workflow, spec versioning rules, impact analysis, or what belongs in specs\u002F vs CLAUDE.md — even if they only ask a question rather than request an edit. Do not use for writing or debugging tests, including contract tests.","zjykzj\u002Fclaude-skills","zjykzj","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F69463217?v=4","01M7BNJKR5W098EGBTX0PPRFHV","01M7BNJKRZHHP642XYS58QJT5H",0,"MIT","https:\u002F\u002Fgithub.com\u002Fzjykzj\u002Fclaude-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fmaestro\u002Fskills\u002Fspec\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: spec\ndescription: Create or modify spec files following project methodology, and drive SDD (spec-first development). Use when writing, editing, or reviewing specs\u002F files; before implementing any feature or behavior change that affects a contract documented in specs\u002F; and whenever the user asks anything about spec-first development — the SDD workflow, spec versioning rules, impact analysis, or what belongs in specs\u002F vs CLAUDE.md — even if they only ask a question rather than request an edit. Do not use for writing or debugging tests, including contract tests.\nallowed-tools: Bash, Read, Write, Edit, Glob, Grep\n---\n\n# Spec Maintenance\n\nApply this methodology when creating or modifying spec files.\n\n## TL;DR\n\n```\nSpec-first → Cross-file scan → Bump version → Implement → Full conformance check → Commit\n```\n\n| What you're doing | How to do it |\n|---|---|\n| Feature \u002F bug fix | Spec first, bump version, then code |\n| **Impact scan (key!)** | `grep` CLAUDE.md, README.md, other specs\u002F for references to the changed contract — add to change list |\n| New definition | Minor bump (v1.0 → v1.1) |\n| Behavioral change (breaking) | Major bump (v1.2 → v2.0) |\n| Clarification \u002F wording | Update date only, keep version |\n| Commit | List **all** affected specs\u002Fdocs in body |\n| Unsure if content belongs | See [curation-guide.md](references\u002Fcuration-guide.md) |\n\n## SDD Workflow (Spec-Driven Development)\n\nAny feature or behavior change that touches a contract documented in `specs\u002F` MUST follow this loop. Specs describe the **target state** — code follows specs, not the other way around.\n\n### Phase 1 — Spec First (before writing any code)\n\n1. **Impact analysis**: determine which spec files\u002Fsections the change affects.\n   If none, the SDD loop does not apply — proceed with normal development.\n2. **Cross-file scan (mandatory)**: search the entire repo for other files that\n   reference the contract being changed — CLAUDE.md, README.md, other specs in\n   `specs\u002F`, and any doc files.  These are also affected and must be updated.\n   Use `grep -rn \"\u003Ckey-term>\" CLAUDE.md README.md specs\u002F` to find ripple effects.\n3. **Update all affected docs**: the primary spec first (bump version per\n   §Version Management), then every cross-referenced file found in step 2.\n4. **Confirm the contract**: present the spec change to the user for approval\n   before implementing.\n\n### Phase 2 — Implement\n\n5. Write code to satisfy the updated spec. If during implementation the contract\n   turns out to be wrong, go back to Phase 1 and change the spec — never silently\n   diverge in code.\n\n### Phase 3 — Conformance Check (after code is complete)\n\n6. **Primary spec**: Re-read every section touched in Phase 1. Verify each\n   contract line against the implementation: signatures, behaviors, edge cases,\n   error\u002Fexit codes.\n7. **Ripple check (mandatory)**: Re-read every cross-referenced file found in\n   Phase 1 step 2.  Verify they are still aligned with the implementation —\n   stale keyboard shortcuts, outdated pipeline descriptions, missing new\n   parameters, etc.\n8. On divergence: code bug → fix the code; contract was wrong → fix the spec\n   (re-bump version), then re-verify.\n\n### Commit Ordering\n\n- Spec and code changes belong to the same commit series; the spec commit\n  **precedes or accompanies** the code commit.\n- The feat\u002Ffix commit body lists **all** affected files — primary spec(s) plus\n  any cross-referenced docs (CLAUDE.md, README.md) found in Phase 1 step 2.\n- Any ripple-fix commits (CLAUDE.md, README.md alignment) discovered during\n  Phase 3 belong in the same series, not deferred to a later session.\n\n## Specs Serve Two Readers\n\n| Reader | Needs from specs |\n|--------|-----------------|\n| **Agent** (AI coding assistant) | Behavioral contracts — \"what is correct\" to verify compliance |\n| **Human developer** | Understanding — \"why\" a contract exists and \"what are the boundaries\" |\n\nBoth readers matter. Content that explains a contract (not just defines it) should be kept.\n\n## Classification Principle\n\nFor each piece of content, ask: **\"When would I read this while writing code?\"**\n\n| Answer | Layer | Typical Content |\n|--------|-------|-----------------|\n| \"Every change\" | **CLAUDE.md** | Architecture hard constraints, global conventions (ordering\u002Fnaming), high-frequency gotchas (encoding rules, state cleanup patterns), critical implementation details that affect every edit |\n| \"Specific task\" | **Specs — WHAT layer** | External contract definitions — what the outside world expects from this system. Data formats, API\u002Fprotocol specs, metric definitions, interface contracts with external systems. Includes explanations that clarify contract semantics (e.g., \"why this field uses center coordinates, not top-left\"). |\n| \"Specific task\" | **Specs — HOW layer** | Internal module contracts — public API signatures, design constraints, option\u002Fparameter definitions, dependency rules, exception types and exit codes. How modules relate to each other, not how they are implemented internally. |\n| \"Neither\" | Delete | — |\n\nWHAT layer naming depends on project domain (see §Spec Directory Structure for the project-type mapping). HOW has exactly one `modules\u002F` layer.\n\n## Content Curation\n\nThe key question: **does this define or help understand a behavioral contract?**\nIf unsure, see the [curation guide](references\u002Fcuration-guide.md) for detailed\njudgment rules (6 content types, pre-deletion checks, interface-vs-implementation\ntable).\n\n## Spec Directory Structure\n\n### WHAT vs HOW Separation\n\n```\nspecs\u002F\n├── \u003Ccontract-layer-1>\u002F    # WHAT — external data\u002Finterface\u002Fprotocol contracts\n│   ├── index.md\n│   └── spec_\u003Ctopic>.md\n│\n├── \u003Ccontract-layer-2>\u002F    # WHAT — other contract layers (optional)\n│   ├── index.md\n│   └── spec_\u003Ctopic>.md\n│\n└── modules\u002F               # HOW — internal module architecture\n    ├── index.md           # Architecture diagram + hard constraints (single source of truth for module dependencies)\n    ├── spec_\u003Cmodule-1>.md\n    ├── spec_\u003Cmodule-2>.md\n    └── ...\n```\n\nWHAT layer naming depends on project domain:\n\n| Project Type | Suggested Name | Example Content |\n|-------------|---------------|-----------------|\n| Data processing \u002F format conversion | `formats\u002F` | Data format definitions, conversion rules |\n| Web API | `api\u002F` | REST\u002FGraphQL interface contracts |\n| Library \u002F SDK | `interfaces\u002F` | Public API signatures, type definitions |\n| Evaluation \u002F benchmarking | `evaluate\u002F` | Metric definitions, baselines |\n| Protocol \u002F communication | `protocols\u002F` | Message formats, state machines |\n\nWHAT layers may have multiple; HOW has exactly one `modules\u002F`, mirroring code modules.\n\n### Index Template\n\nWhen creating a new layer index, use this template:\n\n```markdown\n# \u003CLayer Name> — Specification Index\n\n> **Status:** Canonical — these documents define the authoritative\n> \u003Ccontract-type> for \u003Cproject-name>.\n\n## What This Layer Covers\n\nBriefly describe what this layer defines and what it is the ground truth for.\n\n## Documents\n\n| # | Document | Purpose |\n|---|----------|---------|\n| 1 | `spec_xxx.md` | One-line description |\n\n## Relationship to Other Layers\n\n- This layer (WHAT) maps to which modules in `modules\u002F` (HOW)\n- This layer is independent of which other layers\n\n## Reading Order\n\nRecommended reading order by task:\n- Newcomer → what to read first\n- Specific task → what to read\n```\n\n### What Each Spec Should Answer\n\n| Spec Type | Questions It Answers |\n|-----------|---------------------|\n| Data format \u002F protocol spec | What does this external contract look like? What required fields are defined? |\n| Conversion \u002F adapter spec | How does A become B? How are edge cases handled? |\n| Module spec | What are this module's public interfaces, design constraints, and behavioral contracts? |\n\nA spec file should not answer both \"what does the data look like\" and \"how does the code implement it\" simultaneously. If both appear, split them.\n\n### Version Management\n\nEach spec file starts with a version and last-updated date:\n\n```markdown\n> **Version:** vX.Y | **Last Updated:** YYYY-MM-DD\n```\n\n| Scenario | Version Change |\n|----------|---------------|\n| New definitions \u002F extending existing contracts | Minor increment (v1.0 → v1.1) |\n| Behavioral change (breaking change) | Major increment (v1.2 → v2.0) |\n| Clarification \u002F wording fix (no behavior change) | Update date only, keep version |\n\n## Bootstrap (First Use in a New Project)\n\nThe methodology itself needs no configuration. The SDD enforcement hook ships with the plugin and is active whenever the plugin is installed and enabled — **no project-side hook setup**.\n\nThe only project-side piece is documentation. If the project follows SDD (has a `specs\u002F` directory) but its CLAUDE.md lacks the SDD hard rules, append this section on first use:\n\n```markdown\n## Spec Maintenance\n\n**SDD hard rules:**\n\n1. **Invoke `\u002Fmaestro:spec` before any edit to `specs\u002F` files** — the methodology must be loaded before touching spec content.\n2. **Spec-first ordering**: any feat\u002Ffix that affects a contract documented in `specs\u002F` must (a) update the affected spec to the target state **before** implementing, (b) verify the implementation against the spec **after** coding (conformance check), and (c) list the affected spec files in the commit body.\n```\n\n**Degraded mode**: if the plugin is not installed, there is no hook and no methodology to load — the project's CLAUDE.md rules (above) remain self-sufficient for manual adherence. If the user declines to add the rules, apply the methodology per this skill without persisting anything.\n\n## Required Configuration\n\nNone for the methodology itself — templates are bundled in `templates\u002F` within this skill's directory.\n\nSDD enforcement is provided by the plugin:\n\n- PreToolUse hook (`hooks\u002Fhooks.json`) injects the SDD reminder on `specs\u002F` edits — active automatically while the plugin is enabled\n- CLAUDE.md \"SDD hard rules\" are the project-side documentation (see Bootstrap above)\n","# Spec Maintenance\n\nApply this methodology when creating or modifying spec files.\n\n## TL;DR\n\n```\nSpec-first → Cross-file scan → Bump version → Implement → Full conformance check → Commit\n```\n\n| What you're doing | How to do it |\n|---|---|\n| Feature \u002F bug fix | Spec first, bump version, then code |\n| **Impact scan (key!)** | `grep` CLAUDE.md, README.md, other specs\u002F for references to the changed contract — add to change list |\n| New definition | Minor bump (v1.0 → v1.1) |\n| Behavioral change (breaking) | Major bump (v1.2 → v2.0) |\n| Clarification \u002F wording | Update date only, keep version |\n| Commit | List **all** affected specs\u002Fdocs in body |\n| Unsure if content belongs | See [curation-guide.md](references\u002Fcuration-guide.md) |\n\n## SDD Workflow (Spec-Driven Development)\n\nAny feature or behavior change that touches a contract documented in `specs\u002F` MUST follow this loop. Specs describe the **target state** — code follows specs, not the other way around.\n\n### Phase 1 — Spec First (before writing any code)\n\n1. **Impact analysis**: determine which spec files\u002Fsections the change affects.\n   If none, the SDD loop does not apply — proceed with normal development.\n2. **Cross-file scan (mandatory)**: search the entire repo for other files that\n   reference the contract being changed — CLAUDE.md, README.md, other specs in\n   `specs\u002F`, and any doc files.  These are also affected and must be updated.\n   Use `grep -rn \"\u003Ckey-term>\" CLAUDE.md README.md specs\u002F` to find ripple effects.\n3. **Update all affected docs**: the primary spec first (bump version per\n   §Version Management), then every cross-referenced file found in step 2.\n4. **Confirm the contract**: present the spec change to the user for approval\n   before implementing.\n\n### Phase 2 — Implement\n\n5. Write code to satisfy the updated spec. If during implementation the contract\n   turns out to be wrong, go back to Phase 1 and change the spec — never silently\n   diverge in code.\n\n### Phase 3 — Conformance Check (after code is complete)\n\n6. **Primary spec**: Re-read every section touched in Phase 1. Verify each\n   contract line against the implementation: signatures, behaviors, edge cases,\n   error\u002Fexit codes.\n7. **Ripple check (mandatory)**: Re-read every cross-referenced file found in\n   Phase 1 step 2.  Verify they are still aligned with the implementation —\n   stale keyboard shortcuts, outdated pipeline descriptions, missing new\n   parameters, etc.\n8. On divergence: code bug → fix the code; contract was wrong → fix the spec\n   (re-bump version), then re-verify.\n\n### Commit Ordering\n\n- Spec and code changes belong to the same commit series; the spec commit\n  **precedes or accompanies** the code commit.\n- The feat\u002Ffix commit body lists **all** affected files — primary spec(s) plus\n  any cross-referenced docs (CLAUDE.md, README.md) found in Phase 1 step 2.\n- Any ripple-fix commits (CLAUDE.md, README.md alignment) discovered during\n  Phase 3 belong in the same series, not deferred to a later session.\n\n## Specs Serve Two Readers\n\n| Reader | Needs from specs |\n|--------|-----------------|\n| **Agent** (AI coding assistant) | Behavioral contracts — \"what is correct\" to verify compliance |\n| **Human developer** | Understanding — \"why\" a contract exists and \"what are the boundaries\" |\n\nBoth readers matter. Content that explains a contract (not just defines it) should be kept.\n\n## Classification Principle\n\nFor each piece of content, ask: **\"When would I read this while writing code?\"**\n\n| Answer | Layer | Typical Content |\n|--------|-------|-----------------|\n| \"Every change\" | **CLAUDE.md** | Architecture hard constraints, global conventions (ordering\u002Fnaming), high-frequency gotchas (encoding rules, state cleanup patterns), critical implementation details that affect every edit |\n| \"Specific task\" | **Specs — WHAT layer** | External contract definitions — what the outside world expects from this system. Data formats, API\u002Fprotocol specs, metric definitions, interface contracts with external systems. Includes explanations that clarify contract semantics (e.g., \"why this field uses center coordinates, not top-left\"). |\n| \"Specific task\" | **Specs — HOW layer** | Internal module contracts — public API signatures, design constraints, option\u002Fparameter definitions, dependency rules, exception types and exit codes. How modules relate to each other, not how they are implemented internally. |\n| \"Neither\" | Delete | — |\n\nWHAT layer naming depends on project domain (see §Spec Directory Structure for the project-type mapping). HOW has exactly one `modules\u002F` layer.\n\n## Content Curation\n\nThe key question: **does this define or help understand a behavioral contract?**\nIf unsure, see the [curation guide](references\u002Fcuration-guide.md) for detailed\njudgment rules (6 content types, pre-deletion checks, interface-vs-implementation\ntable).\n\n## Spec Directory Structure\n\n### WHAT vs HOW Separation\n\n```\nspecs\u002F\n├── \u003Ccontract-layer-1>\u002F    # WHAT — external data\u002Finterface\u002Fprotocol contracts\n│   ├── index.md\n│   └── spec_\u003Ctopic>.md\n│\n├── \u003Ccontract-layer-2>\u002F    # WHAT — other contract layers (optional)\n│   ├── index.md\n│   └── spec_\u003Ctopic>.md\n│\n└── modules\u002F               # HOW — internal module architecture\n    ├── index.md           # Architecture diagram + hard constraints (single source of truth for module dependencies)\n    ├── spec_\u003Cmodule-1>.md\n    ├── spec_\u003Cmodule-2>.md\n    └── ...\n```\n\nWHAT layer naming depends on project domain:\n\n| Project Type | Suggested Name | Example Content |\n|-------------|---------------|-----------------|\n| Data processing \u002F format conversion | `formats\u002F` | Data format definitions, conversion rules |\n| Web API | `api\u002F` | REST\u002FGraphQL interface contracts |\n| Library \u002F SDK | `interfaces\u002F` | Public API signatures, type definitions |\n| Evaluation \u002F benchmarking | `evaluate\u002F` | Metric definitions, baselines |\n| Protocol \u002F communication | `protocols\u002F` | Message formats, state machines |\n\nWHAT layers may have multiple; HOW has exactly one `modules\u002F`, mirroring code modules.\n\n### Index Template\n\nWhen creating a new layer index, use this template:\n\n```markdown\n# \u003CLayer Name> — Specification Index\n\n> **Status:** Canonical — these documents define the authoritative\n> \u003Ccontract-type> for \u003Cproject-name>.\n\n## What This Layer Covers\n\nBriefly describe what this layer defines and what it is the ground truth for.\n\n## Documents\n\n| # | Document | Purpose |\n|---|----------|---------|\n| 1 | `spec_xxx.md` | One-line description |\n\n## Relationship to Other Layers\n\n- This layer (WHAT) maps to which modules in `modules\u002F` (HOW)\n- This layer is independent of which other layers\n\n## Reading Order\n\nRecommended reading order by task:\n- Newcomer → what to read first\n- Specific task → what to read\n```\n\n### What Each Spec Should Answer\n\n| Spec Type | Questions It Answers |\n|-----------|---------------------|\n| Data format \u002F protocol spec | What does this external contract look like? What required fields are defined? |\n| Conversion \u002F adapter spec | How does A become B? How are edge cases handled? |\n| Module spec | What are this module's public interfaces, design constraints, and behavioral contracts? |\n\nA spec file should not answer both \"what does the data look like\" and \"how does the code implement it\" simultaneously. If both appear, split them.\n\n### Version Management\n\nEach spec file starts with a version and last-updated date:\n\n```markdown\n> **Version:** vX.Y | **Last Updated:** YYYY-MM-DD\n```\n\n| Scenario | Version Change |\n|----------|---------------|\n| New definitions \u002F extending existing contracts | Minor increment (v1.0 → v1.1) |\n| Behavioral change (breaking change) | Major increment (v1.2 → v2.0) |\n| Clarification \u002F wording fix (no behavior change) | Update date only, keep version |\n\n## Bootstrap (First Use in a New Project)\n\nThe methodology itself needs no configuration. The SDD enforcement hook ships with the plugin and is active whenever the plugin is installed and enabled — **no project-side hook setup**.\n\nThe only project-side piece is documentation. If the project follows SDD (has a `specs\u002F` directory) but its CLAUDE.md lacks the SDD hard rules, append this section on first use:\n\n```markdown\n## Spec Maintenance\n\n**SDD hard rules:**\n\n1. **Invoke `\u002Fmaestro:spec` before any edit to `specs\u002F` files** — the methodology must be loaded before touching spec content.\n2. **Spec-first ordering**: any feat\u002Ffix that affects a contract documented in `specs\u002F` must (a) update the affected spec to the target state **before** implementing, (b) verify the implementation against the spec **after** coding (conformance check), and (c) list the affected spec files in the commit body.\n```\n\n**Degraded mode**: if the plugin is not installed, there is no hook and no methodology to load — the project's CLAUDE.md rules (above) remain self-sufficient for manual adherence. If the user declines to add the rules, apply the methodology per this skill without persisting anything.\n\n## Required Configuration\n\nNone for the methodology itself — templates are bundled in `templates\u002F` within this skill's directory.\n\nSDD enforcement is provided by the plugin:\n\n- PreToolUse hook (`hooks\u002Fhooks.json`) injects the SDD reminder on `specs\u002F` edits — active automatically while the plugin is enabled\n- CLAUDE.md \"SDD hard rules\" are the project-side documentation (see Bootstrap above)\n",{"name":5,"description":7,"allowed-tools":23},"Bash, Read, Write, Edit, Glob, Grep","main","3a702d84c03af43058449a75b3c1999d579c5ff0","plugins\u002Fmaestro\u002Fskills\u002Fspec","https:\u002F\u002Fgithub.com\u002Fzjykzj\u002Fclaude-skills","https:\u002F\u002Fgithub.com\u002Fzjykzj","https:\u002F\u002Fwww.zjykzj.cn","Multimodal & LLM & AIGC","2026-10-25 19:44:12",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add zjykzj\u002Fclaude-skills --skill spec",{"tool":38,"command":39},"bunx","bunx skills add zjykzj\u002Fclaude-skills --skill spec",{"tool":41,"command":42},"pnpm","pnpm dlx skills add zjykzj\u002Fclaude-skills --skill spec","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fzjykzj\u002Fclaude-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fmaestro\u002Fskills\u002Fspec\u002FSKILL.md",true,[46,52,58,62,66],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Fcuration-guide.md","REFERENCE","参考文档",7445,"https:\u002F\u002Fgithub.com\u002Fzjykzj\u002Fclaude-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fmaestro\u002Fskills\u002Fspec\u002Freferences\u002Fcuration-guide.md",{"path":53,"resourceType":54,"resourceTypeDesc":55,"sizeBytes":56,"sourceHtmlUrl":57},"scripts\u002Fsdd-reminder.sh","SCRIPT","脚本",1324,"https:\u002F\u002Fgithub.com\u002Fzjykzj\u002Fclaude-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fmaestro\u002Fskills\u002Fspec\u002Fscripts\u002Fsdd-reminder.sh",{"path":59,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":60,"sourceHtmlUrl":61},"templates\u002Fexamples\u002Fexample_format_spec.md",1596,"https:\u002F\u002Fgithub.com\u002Fzjykzj\u002Fclaude-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fmaestro\u002Fskills\u002Fspec\u002Ftemplates\u002Fexamples\u002Fexample_format_spec.md",{"path":63,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":64,"sourceHtmlUrl":65},"templates\u002Findex_template.md",780,"https:\u002F\u002Fgithub.com\u002Fzjykzj\u002Fclaude-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fmaestro\u002Fskills\u002Fspec\u002Ftemplates\u002Findex_template.md",{"path":67,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":68,"sourceHtmlUrl":69},"templates\u002Fspec_template.md",611,"https:\u002F\u002Fgithub.com\u002Fzjykzj\u002Fclaude-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fmaestro\u002Fskills\u002Fspec\u002Ftemplates\u002Fspec_template.md",[71,77,84,91],{"skillNo":72,"name":73,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":18},"01M7BNK79MT32F0YS1R7SW7NCM","release","https:\u002F\u002Fgithub.com\u002Fzjykzj\u002Fclaude-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fmaestro\u002Fskills\u002Frelease\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":18},"01M7BNK2DX74S6QYJVFQD3Q6YH","commit","Create git commits following project conventions. Use when committing code or preparing a commit message, and whenever the user asks anything about commit conventions — the commit message format, Conventional Commit type classification, the mandatory Co-Authored-By lines, or CHANGELOG maintenance rules (subsections, breaking-change marking, what to skip) — even if they only ask a question rather than request a commit.","https:\u002F\u002Fgithub.com\u002Fzjykzj\u002Fclaude-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fmaestro\u002Fskills\u002Fcommit\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":18},"01M7BNJXQXNSXSRGFH1MEBTNE1","claude","Write or update CLAUDE.md following project conventions. Use when adding gotchas, restructuring CLAUDE.md, or updating project documentation.","https:\u002F\u002Fgithub.com\u002Fzjykzj\u002Fclaude-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fmaestro\u002Fskills\u002Fclaude\u002FSKILL.md",[],[],{"skillNo":92,"name":93,"nameZh":6,"description":94,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":95,"categories":96,"keywords":97,"publishedAt":18},"01M7BNJN61SE3MZ3Y9G89RKYQ5","dataflow-cv","Use the dataflow-cv library (v3.0.0+) — CLI commands (analyse stats\u002Fsplit\u002Ffilter\u002Fpartition\u002Fsample, convert yolo2coco\u002Fyolo2labelme\u002Flabelme2yolo\u002Flabelme2coco\u002Fcoco2yolo\u002Fcoco2labelme, visualize yolo\u002Flabelme\u002Fcoco, evaluate detection\u002Fsegmentation with --prf1) and Python API (label handlers, 5 analysers, 3 converters, 3 visualizers, 2 evaluators, compute_pr_f1, LogConfig\u002FLogManager). Use when working with dataflow-cv — building dataset pipelines, converting YOLO\u002FLabelMe\u002FCOCO annotations, evaluating detections, debugging dataflow-cv errors — and whenever the user asks anything about dataflow-cv — CLI usage, API signatures, metric computation, format conversion rules, or install extras — even if they only ask a question rather than request an action. Do not use for other CV libraries, generic OpenCV usage, or general numpy questions.","https:\u002F\u002Fgithub.com\u002Fzjykzj\u002Fclaude-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fdataflow\u002Fskills\u002Fdataflow-cv\u002FSKILL.md",[],[]]