[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6MMDN9GY7BG6RF1SMPXK385":3,"skill-same:01M6MMDN9GY7BG6RF1SMPXK385":84},{"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":31,"publishedAt":48,"linkOnly":49,"markdownContent":50,"markdownBody":51,"frontMatter":52,"refName":55,"commitSha":56,"skillDirectory":57,"forksCount":58,"repositoryHtmlUrl":59,"repositoryHomepageUrl":60,"ownerHtmlUrl":61,"ownerBlogUrl":62,"ownerDescription":63,"updatedAt":64,"install":65,"resources":77},"01M6MMDN9GY7BG6RF1SMPXK385","tool-design-composition",null,"Compose and expose agent toolsets: task-level operations, batching, discovery, dry runs, and versioning.","hashgraph-online\u002Fawesome-codex-plugins","hashgraph-online","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F160371962?v=4","01M6CNB6NRJVNNGN14TR01Q6D3","HOL (Hashgraph Online)","01M6JT1HSKKG9V68KQAQGBA7XP",1571,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fhashgraph-online\u002Fawesome-codex-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fruntypelabs\u002Fskills\u002Fskills\u002Ftool-design-composition\u002FSKILL.md",[18,23,26],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000015","视觉设计","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",{"categoryNo":27,"categoryName":28,"taxonomyType":29,"taxonomyTypeDesc":30},"01JBSEEDCAT000000000000032","设计师","PROFESSION","适用职业",[32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47],"design","Composition","tool","tool design","tool design composition","SET","discovery","versioning","operations","compose","task","organize","dry","true","batching","extend","2026-11-10 10:12:30",false,"---\nname: tool-design-composition\ndescription: 'Compose and expose agent toolsets: task-level operations, batching, discovery, dry runs, and versioning.'\nuser-invocable: true\nargument-hint: '[tool set to organize or extend]'\n---\n\n# Tool Composition and Discovery Design\n\nA composable tool set has four properties: consistent input and output shapes, so one\ntool's output feeds the next tool's input; batch support, so the agent does not loop one\nitem at a time; more than one abstraction level, so the agent picks the granularity the\ntask needs; and discoverability, so the agent finds the right tool without trial and\nerror.\n\n## Procedure\n\n1. **Map the current set**: every tool, its type, and the sequences agents actually run\n   (from traces, not from intent).\n2. **Find bundle candidates** (Task Bundle): sequences repeated across runs with no\n   branching between steps.\n3. **Find batch candidates** (Batch Operation): the same tool called in a loop over a\n   list.\n4. **Find ladder gaps** (Abstraction Ladder): capabilities with only a raw low-level tool\n   and no intent-level tool, or the reverse.\n5. **Add a `mode` parameter** (Operation Mode) that defaults to `preview` on every\n   destructive or costly command.\n6. **Design discovery** (Dependency Hint, Tool Registry, Schema Explorer): hints in\n   descriptions, a registry or search tool once the set is large, and layered schema\n   exploration for data sources.\n7. **Prune**: merge near-duplicates and retire tools with no traffic. Fewer,\n   better-described tools beat many similar ones.\n8. **Verify**: run each new tool on its own, then run the agent and read its trace to\n   confirm it picks the new bundle, batch form, or preview mode instead of the old\n   sequence.\n\n## Rules with examples\n\n### Offer the capability at more than one altitude (Abstraction Ladder)\n\n```text\nlow   execute_query(sql)                       full control, expert use\nmid   search_records(table, filters)           structured, safe\nhigh  find_customers_like(description)         intent-level, does the work\n```\n\nEach rung has a stated use case; higher rungs call lower ones internally; descriptions\nsay when to use which. Do not ship only the low rung and expect the agent to compose\nit correctly every time.\n\n### Bundle the sequence the agent always runs (Task Bundle)\n\n`search_users` then `open_dm` then `send_message` becomes `dm_user(name, message)`.\nName the bundle after the task, not the steps. Handle intermediate failures inside the\nbundle and return a composite result that reports each step. Keep the atomic tools\navailable for edge cases, and say in the bundle's description that it replaces the\nsequence.\n\n### Accept lists (Batch Operation)\n\nAny tool the agent calls per item over a list gets a batch form: an array input, a\ndocumented maximum batch size, per-item results with status, and a partial-success\nreport (see `tool-design-output`). Batch the upstream calls internally where the API\nallows. Decide up front whether a batch is all-or-nothing or partial, and say which.\n\n### Explore before executing (Operation Mode)\n\nGive commands a `mode` parameter that defaults to the safest option:\n\n```text\nexplore   list what the tool could do here, read-only\npreview   show exactly what would change, no side effects\ndry_run   validate inputs without executing\nexecute   perform the action\n```\n\nReturn the same result structure in every mode so the agent can compare a preview to\nthe real run. A `schema` mode that returns the upstream's complex input shape is useful\nwhen the shape cannot be expressed in the tool interface.\n\n### Make the sequence explicit when it matters (Tool Chain)\n\nFor multi-step processes with a fixed order, define the chain as ordered steps with\ndata passing between them and checkpoints for resume. A chain is a specification the\nagent follows, or a single orchestrated tool that runs it; either way the order is not\nleft to inference.\n\n### Fan out, then merge (Scatter-Gather Tool)\n\nA tool that consults several sources queries them in parallel, merges the results into\none canonical shape, and reports per-source status so a failed source degrades the\nresult rather than failing it.\n\n## Discovery: how the agent finds the right tool\n\n### Say what to call first (Dependency Hint)\n\nDescriptions and parameter docs name prerequisites (\"`userId`: if you only have an\nemail, call `search_users` first\"), follow-ups (\"after this, `get_user` verifies the\nchange\"), and alternatives (\"if this fails, try `send_email`\"). Repeat the hint in the\nerror the tool returns when the prerequisite is missing.\n\n### Catalog the set when it is large (Tool Registry, Capability Matching)\n\nPast roughly twenty tools, tool-selection accuracy drops and every definition costs\ncontext tokens.\nProvide a `list_available_tools(category)` registry with name, description, category,\nand auth requirements, and a `find_tool_by_intent(intent)` search that ranks tools by\nsemantic match with confidence and usage examples. Where your runtime can defer tools,\nkeep the must-use tools always loaded and let the long tail be discovered.\n\n### Reveal structure in layers (Schema Explorer)\n\nFor data sources, do not dump the full schema. Layer it: `list_tables()`, then\n`describe_table(name)`, then `sample_rows(table, limit)`, then `get_query_hints(table)`.\nEach layer's description names the next.\n\n### Check before relying (Health Check)\n\nA fast (sub-second) `check_service_health(service)` returns healthy, degraded, or\nunavailable with specific reasons, so the agent does not spend turns on a dead backend.\nCache the status briefly.\n\n## Across the system\n\n- **Tool Gateway**: one entry point routes to many backends and aggregates discovery,\n  so the agent sees one consistent set.\n- **Tool Adapter**: wrap a legacy or awkward API in a clean tool: LLM-friendly\n  description, constrained inputs, shaped output. The adapter hides the legacy shape\n  entirely.\n- **Canonical Tool Model**: shared `User`, `Task`, `Event` shapes and field names across\n  every tool, mapped to and from each backend.\n- **Tool Versioning**: version in the name (`send_email_v2`) or metadata, run versions in\n  parallel during migration, and deprecate with a migration note in the old tool's\n  description.\n\n## Anti-patterns\n\n- Thirty atomic tools and no bundle for the sequence 80% of runs perform.\n- `delete_files` with no preview mode.\n- Two tools that differ only in one optional parameter.\n- A registry that lists tools without their auth requirements or categories.\n- A composite tool built before any trace showed the sequence being used.\n- A migration that changes a tool's behavior in place instead of versioning it.\n\n## On Runtype\n\n- **Tool count and residency.** Runtype does not cap the number of tools in a dispatch;\n  it caps the request's serialized `runtimeTools` at 1 MB. A model provider can still\n  refuse a long tool list (`TOOL_COUNT_REFUSED_BY_PROVIDER`), so use tool search for\n  large tool sets. Tool search is on by default for multi-turn agents: at 20 or more\n  tools, Runtype keeps a hot set loaded and defers the inline runtime tools behind a\n  `tool_search` tool that the model calls to load more. Saved, built-in, and MCP tools stay loaded. Set\n  `tools.toolSearch.enabled: false` to send every tool on every request. Single-turn\n  agents and flow prompt steps always send every tool, so keep those sets small or split\n  them across subagents. Either way, make descriptions distinct, because they are what\n  the model chooses by and what `tool_search` matches against.\n- **Tool-call budget.** An agent makes at most 10 tool calls per execution unless you\n  set `config.tools.maxToolCalls`. A per-item loop over a list uses up that budget\n  quickly, so ship a batch form, or raise the cap on purpose.\n- **Task bundle and tool chain.** A flow is the platform's explicit chain: fixed step\n  order, data passing, per-step error handling. Expose it as one tool\n  (`toolType: \"flow\"`) when the agent should run the whole sequence as a single call.\n  Map tool parameters to flow inputs with `parameterMapping`. The model receives the\n  flow's terminal step output; to return a composite status object instead, assign it\n  to a flow variable and name that variable in `outputVariable`. A variable name that\n  starts with `_` is rejected, and a variable the flow never assigns fails the call.\n- **Code Mode for batch, chain, and fan-out.** When the logic between calls varies by\n  input (loops over a list, conditional chains, several sources merged), set\n  `config.tools.codeModeConfig` on the agent or prompt step. `toolPool` lists the tool\n  ids the model can call (wildcards such as `mcp:*` work), and `timeoutMs` defaults to 60000. The model writes JavaScript that calls those tools in a sandbox, and\n  intermediate values never enter model context. Choose a `flow` tool when the order\n  is fixed; choose Code Mode when the model must write the logic.\n- **Binary results.** A binary tool result, such as a screenshot or a generated image,\n  carries a `runtype-asset:\u002F\u002F` handle. To accept it in your own tool, declare the input\n  as a string with `\"contentEncoding\": \"base64\"`; a handle passed there is replaced with\n  the stored bytes before your tool runs, so the model never copies base64. A parameter\n  without that declaration receives the handle string unchanged.\n- **Abstraction ladder.** Built-in and Orthogonal catalog tools are the low rung;\n  `custom`, `external`, and `mcp` tools the middle; flows and subagents the\n  top. Compose subagents as saved subagent tools, inline subagent tools, or dynamic\n  spawning through `config.tools.subagentConfig`, which gives the agent a\n  `spawn_subagent` tool. Subagents have their own caps; read\n  `get_platform_documentation(topic=\"subagent-delegation\")`.\n- **Gateway and adapter.** An MCP server attached through `config.tools.mcpServers` is\n  a Tool Gateway: one connection, many tools. An `external` tool is a Tool Adapter: it\n  wraps an upstream API behind a clean interface. For a GraphQL API, use an `external`\n  tool that POSTs the query; the `graphql` tool type is not executable.\n- **Operation mode.** On an agent, gate destructive tools with\n  `config.tools.approval.require` (for example, `[\"delete_files\"]`; patterns such as\n  `mcp:*` work). Approval asks a person before the call runs; it complements a\n  `preview` mode but does not replace it. A `tool-call` flow step runs a catalog tool\n  deterministically with no approval gate, so there a preview must be a distinct tool\n  or a `mode` parameter. Approval details:\n  `get_platform_documentation(topic=\"agent-design\")`.\n- **Description length.** A new saved tool's description is capped at 500 characters. When\n  dependency hints, replacement notes, and deprecation notes do not fit, move them into\n  parameter descriptions or the error text.\n- **Registry and discovery.** `list_tools` covers saved tools,\n  `discover_mcp_server_tools` an MCP server, and\n  `get_platform_documentation(topic=\"builtin-tools\")` the catalog. A `tool-call` step\n  that names a catalog id nobody answers to is rejected as\n  `TOOL_CALL_STEP_UNKNOWN_CATALOG_TOOL`.\n- **Traces and verification.** Find real sequences with `list_runs` (filtered by `agent_id`) and\n  `trace_execution`. Test one tool with `execute_tool`, then run the agent with\n  `execute_agent` and read the trace to confirm the tool choice.\n- **Versioning.** Agents and flows are versioned (`publish_agent_version`,\n  `publish_flow_version`); a tool behavior change ships behind a new published version\n  rather than in place.\n","# Tool Composition and Discovery Design\n\nA composable tool set has four properties: consistent input and output shapes, so one\ntool's output feeds the next tool's input; batch support, so the agent does not loop one\nitem at a time; more than one abstraction level, so the agent picks the granularity the\ntask needs; and discoverability, so the agent finds the right tool without trial and\nerror.\n\n## Procedure\n\n1. **Map the current set**: every tool, its type, and the sequences agents actually run\n   (from traces, not from intent).\n2. **Find bundle candidates** (Task Bundle): sequences repeated across runs with no\n   branching between steps.\n3. **Find batch candidates** (Batch Operation): the same tool called in a loop over a\n   list.\n4. **Find ladder gaps** (Abstraction Ladder): capabilities with only a raw low-level tool\n   and no intent-level tool, or the reverse.\n5. **Add a `mode` parameter** (Operation Mode) that defaults to `preview` on every\n   destructive or costly command.\n6. **Design discovery** (Dependency Hint, Tool Registry, Schema Explorer): hints in\n   descriptions, a registry or search tool once the set is large, and layered schema\n   exploration for data sources.\n7. **Prune**: merge near-duplicates and retire tools with no traffic. Fewer,\n   better-described tools beat many similar ones.\n8. **Verify**: run each new tool on its own, then run the agent and read its trace to\n   confirm it picks the new bundle, batch form, or preview mode instead of the old\n   sequence.\n\n## Rules with examples\n\n### Offer the capability at more than one altitude (Abstraction Ladder)\n\n```text\nlow   execute_query(sql)                       full control, expert use\nmid   search_records(table, filters)           structured, safe\nhigh  find_customers_like(description)         intent-level, does the work\n```\n\nEach rung has a stated use case; higher rungs call lower ones internally; descriptions\nsay when to use which. Do not ship only the low rung and expect the agent to compose\nit correctly every time.\n\n### Bundle the sequence the agent always runs (Task Bundle)\n\n`search_users` then `open_dm` then `send_message` becomes `dm_user(name, message)`.\nName the bundle after the task, not the steps. Handle intermediate failures inside the\nbundle and return a composite result that reports each step. Keep the atomic tools\navailable for edge cases, and say in the bundle's description that it replaces the\nsequence.\n\n### Accept lists (Batch Operation)\n\nAny tool the agent calls per item over a list gets a batch form: an array input, a\ndocumented maximum batch size, per-item results with status, and a partial-success\nreport (see `tool-design-output`). Batch the upstream calls internally where the API\nallows. Decide up front whether a batch is all-or-nothing or partial, and say which.\n\n### Explore before executing (Operation Mode)\n\nGive commands a `mode` parameter that defaults to the safest option:\n\n```text\nexplore   list what the tool could do here, read-only\npreview   show exactly what would change, no side effects\ndry_run   validate inputs without executing\nexecute   perform the action\n```\n\nReturn the same result structure in every mode so the agent can compare a preview to\nthe real run. A `schema` mode that returns the upstream's complex input shape is useful\nwhen the shape cannot be expressed in the tool interface.\n\n### Make the sequence explicit when it matters (Tool Chain)\n\nFor multi-step processes with a fixed order, define the chain as ordered steps with\ndata passing between them and checkpoints for resume. A chain is a specification the\nagent follows, or a single orchestrated tool that runs it; either way the order is not\nleft to inference.\n\n### Fan out, then merge (Scatter-Gather Tool)\n\nA tool that consults several sources queries them in parallel, merges the results into\none canonical shape, and reports per-source status so a failed source degrades the\nresult rather than failing it.\n\n## Discovery: how the agent finds the right tool\n\n### Say what to call first (Dependency Hint)\n\nDescriptions and parameter docs name prerequisites (\"`userId`: if you only have an\nemail, call `search_users` first\"), follow-ups (\"after this, `get_user` verifies the\nchange\"), and alternatives (\"if this fails, try `send_email`\"). Repeat the hint in the\nerror the tool returns when the prerequisite is missing.\n\n### Catalog the set when it is large (Tool Registry, Capability Matching)\n\nPast roughly twenty tools, tool-selection accuracy drops and every definition costs\ncontext tokens.\nProvide a `list_available_tools(category)` registry with name, description, category,\nand auth requirements, and a `find_tool_by_intent(intent)` search that ranks tools by\nsemantic match with confidence and usage examples. Where your runtime can defer tools,\nkeep the must-use tools always loaded and let the long tail be discovered.\n\n### Reveal structure in layers (Schema Explorer)\n\nFor data sources, do not dump the full schema. Layer it: `list_tables()`, then\n`describe_table(name)`, then `sample_rows(table, limit)`, then `get_query_hints(table)`.\nEach layer's description names the next.\n\n### Check before relying (Health Check)\n\nA fast (sub-second) `check_service_health(service)` returns healthy, degraded, or\nunavailable with specific reasons, so the agent does not spend turns on a dead backend.\nCache the status briefly.\n\n## Across the system\n\n- **Tool Gateway**: one entry point routes to many backends and aggregates discovery,\n  so the agent sees one consistent set.\n- **Tool Adapter**: wrap a legacy or awkward API in a clean tool: LLM-friendly\n  description, constrained inputs, shaped output. The adapter hides the legacy shape\n  entirely.\n- **Canonical Tool Model**: shared `User`, `Task`, `Event` shapes and field names across\n  every tool, mapped to and from each backend.\n- **Tool Versioning**: version in the name (`send_email_v2`) or metadata, run versions in\n  parallel during migration, and deprecate with a migration note in the old tool's\n  description.\n\n## Anti-patterns\n\n- Thirty atomic tools and no bundle for the sequence 80% of runs perform.\n- `delete_files` with no preview mode.\n- Two tools that differ only in one optional parameter.\n- A registry that lists tools without their auth requirements or categories.\n- A composite tool built before any trace showed the sequence being used.\n- A migration that changes a tool's behavior in place instead of versioning it.\n\n## On Runtype\n\n- **Tool count and residency.** Runtype does not cap the number of tools in a dispatch;\n  it caps the request's serialized `runtimeTools` at 1 MB. A model provider can still\n  refuse a long tool list (`TOOL_COUNT_REFUSED_BY_PROVIDER`), so use tool search for\n  large tool sets. Tool search is on by default for multi-turn agents: at 20 or more\n  tools, Runtype keeps a hot set loaded and defers the inline runtime tools behind a\n  `tool_search` tool that the model calls to load more. Saved, built-in, and MCP tools stay loaded. Set\n  `tools.toolSearch.enabled: false` to send every tool on every request. Single-turn\n  agents and flow prompt steps always send every tool, so keep those sets small or split\n  them across subagents. Either way, make descriptions distinct, because they are what\n  the model chooses by and what `tool_search` matches against.\n- **Tool-call budget.** An agent makes at most 10 tool calls per execution unless you\n  set `config.tools.maxToolCalls`. A per-item loop over a list uses up that budget\n  quickly, so ship a batch form, or raise the cap on purpose.\n- **Task bundle and tool chain.** A flow is the platform's explicit chain: fixed step\n  order, data passing, per-step error handling. Expose it as one tool\n  (`toolType: \"flow\"`) when the agent should run the whole sequence as a single call.\n  Map tool parameters to flow inputs with `parameterMapping`. The model receives the\n  flow's terminal step output; to return a composite status object instead, assign it\n  to a flow variable and name that variable in `outputVariable`. A variable name that\n  starts with `_` is rejected, and a variable the flow never assigns fails the call.\n- **Code Mode for batch, chain, and fan-out.** When the logic between calls varies by\n  input (loops over a list, conditional chains, several sources merged), set\n  `config.tools.codeModeConfig` on the agent or prompt step. `toolPool` lists the tool\n  ids the model can call (wildcards such as `mcp:*` work), and `timeoutMs` defaults to 60000. The model writes JavaScript that calls those tools in a sandbox, and\n  intermediate values never enter model context. Choose a `flow` tool when the order\n  is fixed; choose Code Mode when the model must write the logic.\n- **Binary results.** A binary tool result, such as a screenshot or a generated image,\n  carries a `runtype-asset:\u002F\u002F` handle. To accept it in your own tool, declare the input\n  as a string with `\"contentEncoding\": \"base64\"`; a handle passed there is replaced with\n  the stored bytes before your tool runs, so the model never copies base64. A parameter\n  without that declaration receives the handle string unchanged.\n- **Abstraction ladder.** Built-in and Orthogonal catalog tools are the low rung;\n  `custom`, `external`, and `mcp` tools the middle; flows and subagents the\n  top. Compose subagents as saved subagent tools, inline subagent tools, or dynamic\n  spawning through `config.tools.subagentConfig`, which gives the agent a\n  `spawn_subagent` tool. Subagents have their own caps; read\n  `get_platform_documentation(topic=\"subagent-delegation\")`.\n- **Gateway and adapter.** An MCP server attached through `config.tools.mcpServers` is\n  a Tool Gateway: one connection, many tools. An `external` tool is a Tool Adapter: it\n  wraps an upstream API behind a clean interface. For a GraphQL API, use an `external`\n  tool that POSTs the query; the `graphql` tool type is not executable.\n- **Operation mode.** On an agent, gate destructive tools with\n  `config.tools.approval.require` (for example, `[\"delete_files\"]`; patterns such as\n  `mcp:*` work). Approval asks a person before the call runs; it complements a\n  `preview` mode but does not replace it. A `tool-call` flow step runs a catalog tool\n  deterministically with no approval gate, so there a preview must be a distinct tool\n  or a `mode` parameter. Approval details:\n  `get_platform_documentation(topic=\"agent-design\")`.\n- **Description length.** A new saved tool's description is capped at 500 characters. When\n  dependency hints, replacement notes, and deprecation notes do not fit, move them into\n  parameter descriptions or the error text.\n- **Registry and discovery.** `list_tools` covers saved tools,\n  `discover_mcp_server_tools` an MCP server, and\n  `get_platform_documentation(topic=\"builtin-tools\")` the catalog. A `tool-call` step\n  that names a catalog id nobody answers to is rejected as\n  `TOOL_CALL_STEP_UNKNOWN_CATALOG_TOOL`.\n- **Traces and verification.** Find real sequences with `list_runs` (filtered by `agent_id`) and\n  `trace_execution`. Test one tool with `execute_tool`, then run the agent with\n  `execute_agent` and read the trace to confirm the tool choice.\n- **Versioning.** Agents and flows are versioned (`publish_agent_version`,\n  `publish_flow_version`); a tool behavior change ships behind a new published version\n  rather than in place.\n",{"name":5,"description":7,"argument-hint":53,"user-invocable":54},"[tool set to organize or extend]",true,"main","2b63f856ce19c2a797929cacabd9e0fc936812fa","plugins\u002Fruntypelabs\u002Fskills\u002Fskills\u002Ftool-design-composition",490,"https:\u002F\u002Fgithub.com\u002Fhashgraph-online\u002Fawesome-codex-plugins","https:\u002F\u002Fhol.org\u002Fplugins\u002Fbest-codex-plugins","https:\u002F\u002Fgithub.com\u002Fhashgraph-online","https:\u002F\u002Fhol.org","Open-source standards and developer tools for interoperable AI agents: discovery, identity, communication, payments, privacy and security.","2026-11-14 02:12:02",{"source":8,"skillName":5,"commands":66,"prompt":76,"zipAvailable":54},[67,70,73],{"tool":68,"command":69},"npx","npx skills add hashgraph-online\u002Fawesome-codex-plugins --skill tool-design-composition",{"tool":71,"command":72},"bunx","bunx skills add hashgraph-online\u002Fawesome-codex-plugins --skill tool-design-composition",{"tool":74,"command":75},"pnpm","pnpm dlx skills add hashgraph-online\u002Fawesome-codex-plugins --skill tool-design-composition","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fhashgraph-online\u002Fawesome-codex-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fruntypelabs\u002Fskills\u002Fskills\u002Ftool-design-composition\u002FSKILL.md",[78],{"path":79,"resourceType":80,"resourceTypeDesc":81,"sizeBytes":82,"sourceHtmlUrl":83},"agents\u002Fopenai.yaml","REFERENCE","参考文档",311,"https:\u002F\u002Fgithub.com\u002Fhashgraph-online\u002Fawesome-codex-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fruntypelabs\u002Fskills\u002Fskills\u002Ftool-design-composition\u002Fagents\u002Fopenai.yaml",[85,93,100,107,133,158],{"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":92},"01M746ZWACVKMZ1BJXJZHYJSSX","migrate-to-e2e","Moves an existing Playwright, Cypress, Selenium, Appium, Detox, Maestro, or TesterArmy suite to e2e one slice at a time, with the old suite running beside the new one until each ported test is proven. Keeps a MIGRATION.md progress file so any later session resumes where the last stopped. Optionally plants a defect to show both suites catch it. Use for \u002Fmigrate-to-e2e, \"migrate our tests to e2e\", \"port these Maestro flows\", \"move our TesterArmy tests\", or when a project has an e2e config and an older suite still running.","https:\u002F\u002Fgithub.com\u002Fhashgraph-online\u002Fawesome-codex-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftester-army\u002Fe2e\u002Fskills\u002Fmigrate-to-e2e\u002FSKILL.md",[],[],"2026-11-11 15:11:14",{"skillNo":94,"name":95,"nameZh":6,"description":96,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":97,"categories":98,"keywords":99,"publishedAt":92},"01M746ZPTAX1Q0FBKAB1CYND7E","e2e","Writes, runs, and debugs end-to-end tests for web and mobile apps with e2e, the agentic test runner, which mixes agent.act goals with exact locator assertions on a Playwright browser or an iOS or Android device. Also drives a project's running app live over the e2e MCP server to look at a screen, check a UI change, take a screenshot or record a video, and try a locator; and runs bug bashes, parallel e2e explore runs with each bug proven by a repro test. Use when a project depends on e2e or has an e2e.config.ts, when asked for end-to-end, browser, mobile, or agentic UI tests, to check a change in the running app, to QA or bug bash an app or branch, or when an e2e run fails.","https:\u002F\u002Fgithub.com\u002Fhashgraph-online\u002Fawesome-codex-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftester-army\u002Fe2e\u002Fskills\u002Fe2e\u002FSKILL.md",[],[],{"skillNo":101,"name":102,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":106,"publishedAt":92},"01M746ZJN784HVHMDP9MQGCSGG","create-verification-skill","Generate a project-local verify-\u003Capp> skill on top of e2e, so any coding agent can launch the app, check it is worth driving, exercise a feature the way a user does, keep evidence, and bug bash it. Writes the skill, a feature map with one explore charter per feature, and proves it once. Use for \u002Fcreate-verification-skill, \"make a verification skill for this repo\", or when a project has no scripted way to prove UI behavior.","https:\u002F\u002Fgithub.com\u002Fhashgraph-online\u002Fawesome-codex-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftester-army\u002Fe2e\u002Fskills\u002Fcreate-verification-skill\u002FSKILL.md",[],[],{"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":113,"publishedAt":48},"01M716A6ETNB9B1SQ3GXZC75VR","deja-search","deja-vu memory — search the user's past AI coding sessions with the deja CLI. Holds what they already settled and never wrote into the repo (naming rules, banned or already-tried tools, required checks before merge, preferred values, who owns what). Use before your first edit in a task, before you call a change done or ready to merge, before you recommend a tool, library, design or reviewer, before re-debugging an error that may already be solved, when they say 'didn't we fix this before', 'what did we decide about X' or 'I already have that', and before telling them something on this machine does not exist.","https:\u002F\u002Fgithub.com\u002Fhashgraph-online\u002Fawesome-codex-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fvshulcz\u002Fdeja-vu\u002Fskills\u002Fdeja-search\u002FSKILL.md",[],[114,115,116,117,118,119,120,121,122,32,123,124,125,126,127,128,129,130,131,132],"search","deja search","never","error","ready","memory","re","decide","fix","debugging","when","you","done","reviewer","io","change","cli","merge","past",{"skillNo":134,"name":135,"nameZh":6,"description":136,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":137,"categories":138,"keywords":139,"publishedAt":48},"01M716A3J2W5SENSE9VGMXB375","deja-reflect","End-of-session reflection and memory curation with deja-vu. Review what was decided, what commands broke and what repaired them, and record durable project rules or decisions with deja remember so the next session starts with them.","https:\u002F\u002Fgithub.com\u002Fhashgraph-online\u002Fawesome-codex-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fvshulcz\u002Fdeja-vu\u002Fskills\u002Fdeja-reflect\u002FSKILL.md",[],[140,141,142,128,143,144,145,146,147,148,149,150,151,152,119,153,154,155,156,157],"deja reflect","reflect","review","with","record","github","next","https","node","Rules","decisions","brew","remember","project rules","project","session","end","reflection",{"skillNo":159,"name":160,"nameZh":6,"description":161,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":162,"categories":163,"keywords":171,"publishedAt":48},"01M715T3RR4BAYSR57HPPTJ0PF","revid","Make, edit, export and post AI videos with Revid (revid.ai) through the Revid MCP server. Use whenever the user wants a short-form or long-form video made (from an idea, a script, a URL, an article, a PDF, a tweet, a product page, their own clips, or a song), wants to find what is going viral on TikTok, YouTube, Instagram or X before making something, wants to edit or re-export an existing Revid project, generate a standalone AI image or clip, make a series episode, or publish and schedule a finished video to connected social accounts. Covers setup (hosted MCP with OAuth, or an API key), the 36 tools, credits, polling and the rules for publishing. Routes to the more specific revid-* skills when one fits.","https:\u002F\u002Fgithub.com\u002Fhashgraph-online\u002Fawesome-codex-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftmakersas\u002Frevid-skills\u002Fskills\u002Frevid\u002FSKILL.md",[164,167,168],{"categoryNo":165,"categoryName":166,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000016","图像与多媒体",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":169,"categoryName":170,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000010","后端与接口",[160,172,173,174,175,176,177,154,178,120,179,180,181,182,124,183,184,185,186,187],"URL","Video","short-form","standalone","PDF","viral","tiktok","short","social","MCP","song","image","mcp server","schedule","api","publish"]