[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M71KCFCNHPE4KZ6VT3V36M8V":3,"skill-same:01M71KCFCNHPE4KZ6VT3V36M8V":81},{"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":23,"publishedAt":44,"linkOnly":45,"markdownContent":46,"markdownBody":47,"frontMatter":48,"refName":51,"commitSha":52,"skillDirectory":53,"forksCount":54,"repositoryHtmlUrl":55,"repositoryHomepageUrl":56,"ownerHtmlUrl":57,"ownerBlogUrl":58,"ownerDescription":59,"updatedAt":60,"install":61,"resources":74},"01M71KCFCNHPE4KZ6VT3V36M8V","curdx-core",null,"Use when handling curdx-flow flags, state files, delegation, execution loops, or skill entrypoint rules.","mugsun\u002Fcurdx-flow","mugsun","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F100790155?v=4","01M6KJSAN2W4WRPDFBKTJTDDQ2","Mugsun","01M71K65VJXA4C1KPJAHZ6ZVHC",4,"MIT","https:\u002F\u002Fgithub.com\u002Fmugsun\u002Fcurdx-flow\u002Fblob\u002Fmain\u002Fplugins\u002Fcurdx-flow\u002Fskills\u002Fcurdx-core\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43],"core","curdx core","state","Rules","use","execution","flow","flags","max","loops","subagent","commit","when","false","loop","coordinator","files","delegation","mode","quick","2026-11-10 10:12:30",false,"---\nname: curdx-core\ndescription: Use when handling curdx-flow flags, state files, delegation, execution loops, or skill entrypoint rules.\nwhen_to_use: Use when user mentions curdx-flow flags, quick mode, commit spec, max iterations, .curdx-state.json, execution loop, coordinator behavior, or subagent delegation.\nversion: 0.2.0\nuser-invocable: false\n---\n\n# curdx-core\n\nCore skill for curdx-flow plugin. Defines common arguments, execution modes, shared behaviors, and coordinator delegation rules.\n\n## Runtime CLI\n\nThe plugin ships a Bash-visible executable: `curdx-flow`.\n\nUse it instead of repeating fragile shell snippets:\n\n```bash\ncurdx-flow route --goal \"$GOAL\" --flags \"$ARGUMENTS\"\ncurdx-flow snapshot --spec \"$SPEC\" --goal \"$GOAL\"\ncurdx-flow snapshot --session-id \"$CLAUDE_SESSION_ID\"\ncurdx-flow specs bind-session \"$SPEC\" --session-id \"$CLAUDE_SESSION_ID\"\ncurdx-flow state merge \"$SPEC_PATH\u002F.curdx-state.json\" '{\"phase\":\"tasks\"}'\ncurdx-flow tasks count \"$SPEC_PATH\u002Ftasks.md\"\ncurdx-flow dev detect\ncurdx-flow dev up\ncurdx-flow dev health\ncurdx-flow dev verify\ncurdx-flow dev down\ncurdx-flow verify run --phase execution --command \"npm test\" --spec \"$SPEC_PATH\"\ncurdx-flow doctor\n```\n\nThe CLI is a thin wrapper around bundled TypeScript helpers under\n`${CLAUDE_PLUGIN_ROOT}\u002Fhooks\u002Fscripts\u002Flib\u002F`. It is the default source of truth\nfor route facts, active spec facts, task counts, and state merge operations.\nWhen Claude Code provides a session id, active spec resolution is session-aware:\nsession binding under `.curdx\u002Fsessions\u002F` wins over the global `.current-spec`\nmarker, while explicit `--spec` always wins over both.\n`curdx-flow verify run` is the preferred way to perform final task\u002Fspec\nverification because it runs the command and records the resulting\n`verificationBlocks.\u003Cphase>` evidence in one step.\n`curdx-flow dev detect\u002Fup\u002Fhealth\u002Fverify\u002Fdown` is the preferred runtime surface\nfor last-mile local evidence. It detects frontend\u002Fbackend roots, starts only\nproject-declared dev commands, checks health endpoints, runs baseline\nverification commands, and stops processes that curdx-flow started.\n`curdx-flow doctor` also reports browser verification readiness: detected E2E\nscripts, Playwright dependencies\u002Fconfig files, chrome-devtools-mcp dependency\ndeclaration, and local Chrome availability.\n\n## Common Arguments\n\nAll curdx-flow public entrypoint skills support these standard arguments:\n\n| Argument | Short | Description | Default |\n|----------|-------|-------------|---------|\n| `--quick` | `-q` | Skip interactive phases, auto-generate artifacts, start execution immediately | false |\n| `--mode` | | Execution policy override: `auto`, `fast`, or `deep` | auto |\n| `--commit-spec` | | Commit and push spec files after each phase | true (normal), false (quick) |\n| `--no-commit-spec` | | Explicitly disable committing spec files | - |\n| `--max-task-iterations` | `-m` | Max retries per failed task before stopping | 5 |\n| `--max-global-iterations` | | Max whole-spec loop iterations before stopping | 30 |\n| `--fresh` | `-f` | Force new spec\u002Ffeature, overwrite if exists | false |\n| `--task-granularity` | | Task granularity override: `auto`, `coarse`, `standard`, or `fine` | auto |\n| `--review` | | Review override: `minimal`, `standard`, or `strict` | policy-derived |\n\nArgument precedence: `--no-commit-spec` > `--commit-spec` > mode default.\n\n## Execution Modes\n\n### Auto Policy (Default)\n\nEvery new spec runs deterministic policy classification before expensive model work:\n\n```bash\ncurdx-flow route --goal \"$GOAL\" --flags \"$ARGUMENTS\"\n```\n\nPersist the JSON as `.curdx-state.json::autoPolicy`. Later phases must obey it\ninstead of asking the user to choose fast\u002Fdeep. Policy controls:\n- behavior-route compatibility fields\n- risk level\n- task target range\n- review cadence\n- verification level\n- subagent policy\n- native `\u002Fgoal` turn driver and deterministic Stop-hook gates\n\n### Normal Mode (Interactive)\n\n- User reviews artifacts between phases\n- Phase transitions require explicit `\u002Fcurdx-flow:*` skill invocation\n- Each phase sets `awaitingApproval: true`\n- Commits spec files by default\n\n### Quick Mode (`--quick`)\n\n- Skip all interactive prompts, interviews, and approval pauses\n- Run the same phase agents (research, requirements, design, tasks) sequentially\n- Agents receive a \"be more opinionated\" directive since there is no user feedback\n- spec-reviewer validates each artifact (max 3 iterations)\n- Immediately start execution after all phases complete\n- Do NOT commit spec files by default (use `--commit-spec` to override)\n- Still delegate to subagents (delegation is mandatory)\n\n## State File\n\ncurdx-flow uses `.curdx-state.json` for execution state. New state files use\n`version: 2`; legacy state may be reinitialized rather than migrated. See\n`references\u002Fstate-file-schema.md` for the current schema.\n\nKey fields: `phase`, `taskIndex`, `totalTasks`, `taskIteration`, `maxTaskIterations`, `awaitingApproval`.\n\nPhase skills must start by running `curdx-flow snapshot`. Treat the snapshot's\n`gates` list as blocking unless the current skill explicitly owns that gate.\nSnapshots also carry recovery facts from `.curdx\u002Fbrain.jsonl`: recent verifier\nfailures and the last official PostCompact summary. Hooks inject only compact\nCURDX SPEC DATA capsules with pointers and state; they do not inject full\nartifact bodies as instructions.\n\n## Commit Behavior\n\nWhen `commitSpec` is true:\n\n1. Stage spec\u002Ffeature files after generation\n2. Commit with message: `chore(\u003Cplugin>): commit spec files before implementation`\n3. Push to current branch\n\nWhen `commitSpec` is false:\n\n- Files remain uncommitted\n- User can manually commit later\n\n## Task Execution Loop\n\ncurdx-flow v3.0.0+ has a self-contained execution loop via the stop-hook. Companion plugins and MCPs improve routing, memory, docs, UI, browser proof, reasoning, and recovery, but the stop-hook loop itself does not delegate to an external orchestrator.\n\nKey signals:\n- `TASK_COMPLETE` - executor finished task\n- `ALL_TASKS_COMPLETE` - coordinator ends loop\n\n## Error Handling\n\nWhen `taskIteration > maxTaskIterations`: block task, suggest manual intervention.\n\nIf state file missing\u002Finvalid: output error, suggest re-running `\u002Fcurdx-flow:implement`.\n\n## Branch Management\n\ncurdx-flow follows a consistent branch strategy:\n\n1. Check current branch before starting\n2. If on default branch (main\u002Fmaster): prompt for branch strategy\n3. If on feature branch: offer to continue or create new\n4. Quick mode: auto-create branch, no prompts\n\n## Coordinator Behavior\n\nThe main agent is a coordinator, not an implementer. Delegate all work to subagents.\n\n### Fan-Out Heuristic\n\nThe coordinator's default disposition is **fan out unless a cheaper path is provably sufficient**. This implements Anthropic's canonical guidance — *\"Use [a subagent] when a side task would flood your main conversation with search results, logs, or file contents you won't reference again\"* (`code.claude.com\u002Fdocs\u002Fen\u002Fsub-agents`).\n\n**Fan out by default when ANY of these hold:**\n\n1. **Context pollution** — the work would read >3 files or >2 web pages whose raw content the main context won't reference again. Single-agent fan-out is correct usage; parallelism is a bonus, not a precondition.\n2. **Independent multi-topic** — ≥2 topics where independent input \u002F independent output \u002F independent context (the 3 criteria in `${CLAUDE_PLUGIN_ROOT}\u002Freferences\u002Fbounded-parallel-dispatch.md`) all hold.\n3. **Downstream synthesis** — findings will feed an artifact, interview question, or decision the coordinator owns.\n\n**Do NOT fan out when:**\n\n- Exact string match \u002F known glob \u002F single deterministic edit — `grep` and `Edit` beat 30-60s subagent overhead.\n- The route output already gives the answer (don't re-derive what `curdx-flow route` told you).\n- Debug investigation BEFORE sequential triage narrowed to ≤3 leads (anti-pattern #12, debug domain).\n- You're tempted to use a Discovery agent to write code or commit files — Discovery is read-only (anti-pattern #18).\n\n**Hard rules when you do fan out:**\n\n- ALL `Agent(...)` calls in ONE message for true parallelism (anti-pattern #3, research domain).\n- Each agent gets a self-contained brief: goal, context, output cap, scope fence (see `${CLAUDE_PLUGIN_ROOT}\u002Freferences\u002Fprompt-optimization.md` and `${CLAUDE_PLUGIN_ROOT}\u002Freferences\u002Fagent-output-contract.md`).\n- Coordinator synthesizes; coordinator does not parallel-do-the-same-work it delegated (system-prompt rule: *\"avoid duplicating work that subagents are already doing\"*).\n- Trust but verify — an agent's summary describes intent, not necessarily reality. Independently verify load-bearing claims before acting on them.\n\n**Pre-spec coordinator default** (the most common gap):\n\nFor `product-inception`, `greenfield-spec`, `prototype`, `import-spec`, or `blocked-ask-user` with non-empty `intent.missingFacts`, fan out a **Discovery** batch (1 claude-mem search + 1 Explore at `.planning\u002F`\u002F`docs\u002F`\u002F`CLAUDE.md` + N research-analyst, one per candidate stack or prior-art topic) BEFORE asking the user any question or writing any product-context artifact. See `${CLAUDE_PLUGIN_ROOT}\u002Freferences\u002Fbounded-parallel-dispatch.md` Discovery domain and `${CLAUDE_PLUGIN_ROOT}\u002Fskills\u002Fstart\u002FSKILL.md` Pre-Question Discovery section.\n\n## Skill Entrypoints\n\ncurdx-flow is skills-only at the plugin surface.\n\n- Public `\u002Fcurdx-flow:*` entries live in `skills\u002F\u003Cname>\u002FSKILL.md`.\n- Shared\u002Fbackground behavior lives in support skills such as `spec-workflow`, `curdx-core`, and `interview-framework`.\n- Keep phase-changing entry skills explicit with `disable-model-invocation: true`; let support skills handle automatic guidance.\n\n### Coordinator Responsibilities\n\n1. Parse user input and determine intent\n2. Read state files for context\n3. Delegate work to subagents via Agent tool\n4. Report results to user\n\n### Do Not\n\n- Write code, create files, or modify source directly\n- Run implementation commands (npm, git commit, file edits)\n- Perform research, analysis, or design directly\n- Execute task steps from tasks.md\n\n### Delegation Mapping\n\n| Work Type | Delegate To |\n|-----------|-------------|\n| Research | Bounded parallel direct `Agent(...)` dispatch to `research-analyst` \u002F Explore-style research |\n| Requirements | product-manager subagent |\n| Design | architect-reviewer subagent |\n| Task planning | task-planner subagent |\n| Task execution | spec-executor subagent |\n\nQuick mode still requires delegation.\n","# curdx-core\n\nCore skill for curdx-flow plugin. Defines common arguments, execution modes, shared behaviors, and coordinator delegation rules.\n\n## Runtime CLI\n\nThe plugin ships a Bash-visible executable: `curdx-flow`.\n\nUse it instead of repeating fragile shell snippets:\n\n```bash\ncurdx-flow route --goal \"$GOAL\" --flags \"$ARGUMENTS\"\ncurdx-flow snapshot --spec \"$SPEC\" --goal \"$GOAL\"\ncurdx-flow snapshot --session-id \"$CLAUDE_SESSION_ID\"\ncurdx-flow specs bind-session \"$SPEC\" --session-id \"$CLAUDE_SESSION_ID\"\ncurdx-flow state merge \"$SPEC_PATH\u002F.curdx-state.json\" '{\"phase\":\"tasks\"}'\ncurdx-flow tasks count \"$SPEC_PATH\u002Ftasks.md\"\ncurdx-flow dev detect\ncurdx-flow dev up\ncurdx-flow dev health\ncurdx-flow dev verify\ncurdx-flow dev down\ncurdx-flow verify run --phase execution --command \"npm test\" --spec \"$SPEC_PATH\"\ncurdx-flow doctor\n```\n\nThe CLI is a thin wrapper around bundled TypeScript helpers under\n`${CLAUDE_PLUGIN_ROOT}\u002Fhooks\u002Fscripts\u002Flib\u002F`. It is the default source of truth\nfor route facts, active spec facts, task counts, and state merge operations.\nWhen Claude Code provides a session id, active spec resolution is session-aware:\nsession binding under `.curdx\u002Fsessions\u002F` wins over the global `.current-spec`\nmarker, while explicit `--spec` always wins over both.\n`curdx-flow verify run` is the preferred way to perform final task\u002Fspec\nverification because it runs the command and records the resulting\n`verificationBlocks.\u003Cphase>` evidence in one step.\n`curdx-flow dev detect\u002Fup\u002Fhealth\u002Fverify\u002Fdown` is the preferred runtime surface\nfor last-mile local evidence. It detects frontend\u002Fbackend roots, starts only\nproject-declared dev commands, checks health endpoints, runs baseline\nverification commands, and stops processes that curdx-flow started.\n`curdx-flow doctor` also reports browser verification readiness: detected E2E\nscripts, Playwright dependencies\u002Fconfig files, chrome-devtools-mcp dependency\ndeclaration, and local Chrome availability.\n\n## Common Arguments\n\nAll curdx-flow public entrypoint skills support these standard arguments:\n\n| Argument | Short | Description | Default |\n|----------|-------|-------------|---------|\n| `--quick` | `-q` | Skip interactive phases, auto-generate artifacts, start execution immediately | false |\n| `--mode` | | Execution policy override: `auto`, `fast`, or `deep` | auto |\n| `--commit-spec` | | Commit and push spec files after each phase | true (normal), false (quick) |\n| `--no-commit-spec` | | Explicitly disable committing spec files | - |\n| `--max-task-iterations` | `-m` | Max retries per failed task before stopping | 5 |\n| `--max-global-iterations` | | Max whole-spec loop iterations before stopping | 30 |\n| `--fresh` | `-f` | Force new spec\u002Ffeature, overwrite if exists | false |\n| `--task-granularity` | | Task granularity override: `auto`, `coarse`, `standard`, or `fine` | auto |\n| `--review` | | Review override: `minimal`, `standard`, or `strict` | policy-derived |\n\nArgument precedence: `--no-commit-spec` > `--commit-spec` > mode default.\n\n## Execution Modes\n\n### Auto Policy (Default)\n\nEvery new spec runs deterministic policy classification before expensive model work:\n\n```bash\ncurdx-flow route --goal \"$GOAL\" --flags \"$ARGUMENTS\"\n```\n\nPersist the JSON as `.curdx-state.json::autoPolicy`. Later phases must obey it\ninstead of asking the user to choose fast\u002Fdeep. Policy controls:\n- behavior-route compatibility fields\n- risk level\n- task target range\n- review cadence\n- verification level\n- subagent policy\n- native `\u002Fgoal` turn driver and deterministic Stop-hook gates\n\n### Normal Mode (Interactive)\n\n- User reviews artifacts between phases\n- Phase transitions require explicit `\u002Fcurdx-flow:*` skill invocation\n- Each phase sets `awaitingApproval: true`\n- Commits spec files by default\n\n### Quick Mode (`--quick`)\n\n- Skip all interactive prompts, interviews, and approval pauses\n- Run the same phase agents (research, requirements, design, tasks) sequentially\n- Agents receive a \"be more opinionated\" directive since there is no user feedback\n- spec-reviewer validates each artifact (max 3 iterations)\n- Immediately start execution after all phases complete\n- Do NOT commit spec files by default (use `--commit-spec` to override)\n- Still delegate to subagents (delegation is mandatory)\n\n## State File\n\ncurdx-flow uses `.curdx-state.json` for execution state. New state files use\n`version: 2`; legacy state may be reinitialized rather than migrated. See\n`references\u002Fstate-file-schema.md` for the current schema.\n\nKey fields: `phase`, `taskIndex`, `totalTasks`, `taskIteration`, `maxTaskIterations`, `awaitingApproval`.\n\nPhase skills must start by running `curdx-flow snapshot`. Treat the snapshot's\n`gates` list as blocking unless the current skill explicitly owns that gate.\nSnapshots also carry recovery facts from `.curdx\u002Fbrain.jsonl`: recent verifier\nfailures and the last official PostCompact summary. Hooks inject only compact\nCURDX SPEC DATA capsules with pointers and state; they do not inject full\nartifact bodies as instructions.\n\n## Commit Behavior\n\nWhen `commitSpec` is true:\n\n1. Stage spec\u002Ffeature files after generation\n2. Commit with message: `chore(\u003Cplugin>): commit spec files before implementation`\n3. Push to current branch\n\nWhen `commitSpec` is false:\n\n- Files remain uncommitted\n- User can manually commit later\n\n## Task Execution Loop\n\ncurdx-flow v3.0.0+ has a self-contained execution loop via the stop-hook. Companion plugins and MCPs improve routing, memory, docs, UI, browser proof, reasoning, and recovery, but the stop-hook loop itself does not delegate to an external orchestrator.\n\nKey signals:\n- `TASK_COMPLETE` - executor finished task\n- `ALL_TASKS_COMPLETE` - coordinator ends loop\n\n## Error Handling\n\nWhen `taskIteration > maxTaskIterations`: block task, suggest manual intervention.\n\nIf state file missing\u002Finvalid: output error, suggest re-running `\u002Fcurdx-flow:implement`.\n\n## Branch Management\n\ncurdx-flow follows a consistent branch strategy:\n\n1. Check current branch before starting\n2. If on default branch (main\u002Fmaster): prompt for branch strategy\n3. If on feature branch: offer to continue or create new\n4. Quick mode: auto-create branch, no prompts\n\n## Coordinator Behavior\n\nThe main agent is a coordinator, not an implementer. Delegate all work to subagents.\n\n### Fan-Out Heuristic\n\nThe coordinator's default disposition is **fan out unless a cheaper path is provably sufficient**. This implements Anthropic's canonical guidance — *\"Use [a subagent] when a side task would flood your main conversation with search results, logs, or file contents you won't reference again\"* (`code.claude.com\u002Fdocs\u002Fen\u002Fsub-agents`).\n\n**Fan out by default when ANY of these hold:**\n\n1. **Context pollution** — the work would read >3 files or >2 web pages whose raw content the main context won't reference again. Single-agent fan-out is correct usage; parallelism is a bonus, not a precondition.\n2. **Independent multi-topic** — ≥2 topics where independent input \u002F independent output \u002F independent context (the 3 criteria in `${CLAUDE_PLUGIN_ROOT}\u002Freferences\u002Fbounded-parallel-dispatch.md`) all hold.\n3. **Downstream synthesis** — findings will feed an artifact, interview question, or decision the coordinator owns.\n\n**Do NOT fan out when:**\n\n- Exact string match \u002F known glob \u002F single deterministic edit — `grep` and `Edit` beat 30-60s subagent overhead.\n- The route output already gives the answer (don't re-derive what `curdx-flow route` told you).\n- Debug investigation BEFORE sequential triage narrowed to ≤3 leads (anti-pattern #12, debug domain).\n- You're tempted to use a Discovery agent to write code or commit files — Discovery is read-only (anti-pattern #18).\n\n**Hard rules when you do fan out:**\n\n- ALL `Agent(...)` calls in ONE message for true parallelism (anti-pattern #3, research domain).\n- Each agent gets a self-contained brief: goal, context, output cap, scope fence (see `${CLAUDE_PLUGIN_ROOT}\u002Freferences\u002Fprompt-optimization.md` and `${CLAUDE_PLUGIN_ROOT}\u002Freferences\u002Fagent-output-contract.md`).\n- Coordinator synthesizes; coordinator does not parallel-do-the-same-work it delegated (system-prompt rule: *\"avoid duplicating work that subagents are already doing\"*).\n- Trust but verify — an agent's summary describes intent, not necessarily reality. Independently verify load-bearing claims before acting on them.\n\n**Pre-spec coordinator default** (the most common gap):\n\nFor `product-inception`, `greenfield-spec`, `prototype`, `import-spec`, or `blocked-ask-user` with non-empty `intent.missingFacts`, fan out a **Discovery** batch (1 claude-mem search + 1 Explore at `.planning\u002F`\u002F`docs\u002F`\u002F`CLAUDE.md` + N research-analyst, one per candidate stack or prior-art topic) BEFORE asking the user any question or writing any product-context artifact. See `${CLAUDE_PLUGIN_ROOT}\u002Freferences\u002Fbounded-parallel-dispatch.md` Discovery domain and `${CLAUDE_PLUGIN_ROOT}\u002Fskills\u002Fstart\u002FSKILL.md` Pre-Question Discovery section.\n\n## Skill Entrypoints\n\ncurdx-flow is skills-only at the plugin surface.\n\n- Public `\u002Fcurdx-flow:*` entries live in `skills\u002F\u003Cname>\u002FSKILL.md`.\n- Shared\u002Fbackground behavior lives in support skills such as `spec-workflow`, `curdx-core`, and `interview-framework`.\n- Keep phase-changing entry skills explicit with `disable-model-invocation: true`; let support skills handle automatic guidance.\n\n### Coordinator Responsibilities\n\n1. Parse user input and determine intent\n2. Read state files for context\n3. Delegate work to subagents via Agent tool\n4. Report results to user\n\n### Do Not\n\n- Write code, create files, or modify source directly\n- Run implementation commands (npm, git commit, file edits)\n- Perform research, analysis, or design directly\n- Execute task steps from tasks.md\n\n### Delegation Mapping\n\n| Work Type | Delegate To |\n|-----------|-------------|\n| Research | Bounded parallel direct `Agent(...)` dispatch to `research-analyst` \u002F Explore-style research |\n| Requirements | product-manager subagent |\n| Design | architect-reviewer subagent |\n| Task planning | task-planner subagent |\n| Task execution | spec-executor subagent |\n\nQuick mode still requires delegation.\n",{"name":5,"version":49,"description":7,"when_to_use":50,"user-invocable":45},"0.2.0","Use when user mentions curdx-flow flags, quick mode, commit spec, max iterations, .curdx-state.json, execution loop, coordinator behavior, or subagent delegation.","main","7357746802ae7a84cafa93b9dbffd138817b0290","plugins\u002Fcurdx-flow\u002Fskills\u002Fcurdx-core",0,"https:\u002F\u002Fgithub.com\u002Fmugsun\u002Fcurdx-flow","https:\u002F\u002Fcurdx.org","https:\u002F\u002Fgithub.com\u002Fmugsun","📮 bydongxin@gmail.com","🧑‍💻 AI Builder & Software Engineer\r\n🚀 Turning ideas into software.          \r\n\r\n","2026-08-19 15:19:29",{"source":8,"skillName":5,"commands":62,"prompt":72,"zipAvailable":73},[63,66,69],{"tool":64,"command":65},"npx","npx skills add mugsun\u002Fcurdx-flow --skill curdx-core",{"tool":67,"command":68},"bunx","bunx skills add mugsun\u002Fcurdx-flow --skill curdx-core",{"tool":70,"command":71},"pnpm","pnpm dlx skills add mugsun\u002Fcurdx-flow --skill curdx-core","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fmugsun\u002Fcurdx-flow\u002Fblob\u002Fmain\u002Fplugins\u002Fcurdx-flow\u002Fskills\u002Fcurdx-core\u002FSKILL.md",true,[75],{"path":76,"resourceType":77,"resourceTypeDesc":78,"sizeBytes":79,"sourceHtmlUrl":80},"references\u002Fstate-file-schema.md","REFERENCE","参考文档",7023,"https:\u002F\u002Fgithub.com\u002Fmugsun\u002Fcurdx-flow\u002Fblob\u002Fmain\u002Fplugins\u002Fcurdx-flow\u002Fskills\u002Fcurdx-core\u002Freferences\u002Fstate-file-schema.md",[82,91,107,124,138,149],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":89,"publishedAt":44},"01M71KDR5EJP6G4VWFCHNJ35PT","help","Fixture help skill","https:\u002F\u002Fgithub.com\u002Fmugsun\u002Fcurdx-flow\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fruntime-readiness\u002Fclaude-code-plugin-like\u002Fskills\u002Fhelp\u002FSKILL.md",[88],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[84,90],"fixture",{"skillNo":92,"name":93,"nameZh":6,"description":94,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":95,"categories":96,"keywords":97,"publishedAt":44},"01M71KDNAVB000W3MZB1W7CMD8","verification-before-completion","Use when checking fixes, phase exits, completion evidence, or mock-heavy tests before claiming success.","https:\u002F\u002Fgithub.com\u002Fmugsun\u002Fcurdx-flow\u002Fblob\u002Fmain\u002Fplugins\u002Fcurdx-flow\u002Fskills\u002Fverification-before-completion\u002FSKILL.md",[],[98,99,100,37,36,101,102,103,104,105,28,106,26],"verification","verification before completion","only","tasks","phase","evidence","fixes","tests","mock",{"skillNo":108,"name":109,"nameZh":6,"description":110,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":111,"categories":112,"keywords":114,"publishedAt":44},"01M71KDKFE975GFYRDC21H1NEE","triage","Use when a large feature should be decomposed into dependency-aware specs.","https:\u002F\u002Fgithub.com\u002Fmugsun\u002Fcurdx-flow\u002Fblob\u002Fmain\u002Fplugins\u002Fcurdx-flow\u002Fskills\u002Ftriage\u002FSKILL.md",[113],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[109,115,116,117,118,119,120,121,122,36,28,123],"goal","true","feature","write","read","dependency","name","Bash","EPIC",{"skillNo":125,"name":101,"nameZh":6,"description":126,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":127,"categories":128,"keywords":130,"publishedAt":44},"01M71KDH7XFQJPKYZ0H9JDNSA5","Use when a spec has design.md and needs implementation tasks.","https:\u002F\u002Fgithub.com\u002Fmugsun\u002Fcurdx-flow\u002Fblob\u002Fmain\u002Fplugins\u002Fcurdx-flow\u002Fskills\u002Ftasks\u002FSKILL.md",[129],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[101,131,132,121,133,28,134,135,136,36,122,119,118,137,116],"implementation","spec","task","standard","design","auto","edit",{"skillNo":139,"name":140,"nameZh":6,"description":141,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":142,"categories":143,"keywords":145,"publishedAt":44},"01M71KDEV9G51MCXDQWPJH7QST","switch","Use when changing the active spec by name or path.","https:\u002F\u002Fgithub.com\u002Fmugsun\u002Fcurdx-flow\u002Fblob\u002Fmain\u002Fplugins\u002Fcurdx-flow\u002Fskills\u002Fswitch\u002FSKILL.md",[144],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[140,122,146,116,28,36,147,121,119,132,118,148],"by name","active","path",{"skillNo":150,"name":151,"nameZh":6,"description":152,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":153,"categories":154,"keywords":156,"publishedAt":44},"01M71KDDJFRV09YZ08B3WJ39SC","status","Use when inspecting curdx-flow specs, active state, progress, or index status.","https:\u002F\u002Fgithub.com\u002Fmugsun\u002Fcurdx-flow\u002Fblob\u002Fmain\u002Fplugins\u002Fcurdx-flow\u002Fskills\u002Fstatus\u002FSKILL.md",[155],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[151,116,28,36,122,26,157,147,158,119,30,159],"progress","update","index"]