[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7F3TQPJ2DG1H5DHNTRSVASH":3,"skill-same:01M7F3TQPJ2DG1H5DHNTRSVASH":69},{"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":30,"commitSha":31,"skillDirectory":32,"forksCount":33,"repositoryHtmlUrl":34,"repositoryHomepageUrl":6,"ownerHtmlUrl":35,"ownerBlogUrl":6,"ownerDescription":36,"updatedAt":37,"install":38,"resources":50},"01M7F3TQPJ2DG1H5DHNTRSVASH","rd-agent",null,"Operate Microsoft's RD-Agent research-agent framework for data-science, quantitative-finance, LLM fine-tuning, reinforcement-learning, and paper\u002Fmodel-copilot workflows.","lza6\u002FArmor-piercing-system-prompts","lza6","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F128880206?v=4","01M7EB3K7WPX5T28EJ2NVHE6EG","Chinese-tingfeng","01M7EB3K97E1KQ16QWKJMFNGX8",11,"MIT","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F05-技能与代理开发\u002Frd-agent\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: rd-agent\ndescription: \"Operate Microsoft's RD-Agent research-agent framework for\n  data-science, quantitative-finance, LLM fine-tuning, reinforcement-learning,\n  and paper\u002Fmodel-copilot workflows.\"\nmetadata:\n  disco-role: operating\n  source-repository: microsoft\u002FRD-Agent\n  source-revision: 6762f84f9bc0f5c6486c50a00e128a57ac6c3683\n  package: rdagent\nlicense: MIT\ndisable-model-invocation: true\n---\n\n# RD-Agent\n\nUse this skill when a task involves the **RD-Agent** Python package or a checkout of the `microsoft\u002FRD-Agent` repository. RD-Agent is an agentic research framework, not a single model: it composes LLM calls, code generation, execution, evaluation, and experiment logs around several scenario families.\n\n## Route first\n\nChoose the smallest sub-skill that covers the user's requested workflow:\n\n| Request shape | Read next |\n|---|---|\n| Install, environment checks, CLI help, UI\u002Fserver startup, logs, config discovery, or safe maintenance | [setup-and-ops](sub-skills\u002Fsetup-and-ops\u002FSKILL.md) |\n| Qlib factors, factor reports, quantitative research, backtests, or finance data-agent\u002Fmodel-agent flows | [quant-finance](sub-skills\u002Fquant-finance\u002FSKILL.md) |\n| Kaggle-style tabular work, data-science agent loops, dataset preparation, or competition artifacts | [competition-data-science](sub-skills\u002Fcompetition-data-science\u002FSKILL.md) |\n| LLM fine-tuning datasets, training jobs, merge\u002Fevaluation, or fine-tune UI | [llm-finetune](sub-skills\u002Fllm-finetune\u002FSKILL.md) |\n| AutoRL-Bench, RL scenario loops, benchmark agents, or post-training experiments | [rl-post-training](sub-skills\u002Frl-post-training\u002FSKILL.md) |\n| General model copilot, paper-driven model exploration, or reusable model-research scaffolding | [paper-model-copilot](sub-skills\u002Fpaper-model-copilot\u002FSKILL.md) |\n\nFor a mixed request, read the relevant sub-skills in the order **setup → domain workflow → evaluation\u002Ftroubleshooting**. Do not load every sub-skill by default.\n\n## Operating model\n\n1. **Establish the runtime.** Confirm Python is supported (the inspected project supports Python 3.10 and 3.11), the package is importable, and the requested command is available. For a checkout, install with `python -m pip install -e .`; for an installed distribution, use the package's normal environment. Keep API keys, data roots, model caches, and experiment outputs outside the skill tree.\n2. **Select the scenario.** RD-Agent's public CLI exposes `data_science`, `fin_quant`, `fin_factor_report`, `llm_finetune`, `general_model`, `ui`, `server_ui`, and `ds_user_interact`. The RL benchmark entry point is `python -m rdagent.scenarios.rl.autorl_bench.run`.\n3. **Make configuration explicit.** Prefer an existing example or scenario config, then override only the values needed for the task. Record dataset paths, model\u002Fprovider settings, execution sandbox, output directory, and evaluation metric before launching an agent loop.\n4. **Run the smallest smoke check.** Use `--help`, import checks, or a tiny fixture before a long experiment. Do not infer a successful experiment from process startup: inspect the generated summary, evaluator result, and logs.\n5. **Preserve evidence.** Save the command, resolved configuration, commit\u002Fpackage version, metric definitions, and failure traceback with the experiment. Keep generated code and datasets in a separate run directory.\n\n## Common preflight\n\n```bash\npython -c \"import rdagent; print(rdagent.__file__)\"\nrdagent --help\nrdagent health_check --no-check-env --no-check-docker\n```\n\nThe health check can inspect Docker and external tools when those checks are enabled. Use the explicit `--no-check-*` flags only for a lightweight package\u002FCLI smoke test; do not treat that result as proof that a containerized or GPU-backed workflow is ready.\n\n## Safety and scope boundaries\n\n- Agent loops can generate and execute code. Start with a disposable output directory and a restricted execution environment.\n- Finance, Kaggle, model-training, and RL scenarios may require credentials, large datasets, Docker, CUDA, or external services. State those prerequisites instead of silently substituting a CPU or toy result.\n- Avoid launching full training, benchmark servers, UI processes, or network downloads as a generic smoke test. Use the bundled helper scripts for deterministic checks.\n- `fitz` may emit a deprecation warning: use of the warning alone is not a failure. AutoRL-Bench may report an empty registry when `SMITH_BENCH_DIR` is unset or points to a missing benchmark checkout.\n- Never place provider keys, cookies, competition credentials, private data, or local absolute paths in a generated report or a reusable configuration.\n\n## Evidence map\n\n- [Package and CLI overview](references\u002Fpackage-overview.md) — public entry points and component boundaries.\n- [Troubleshooting](references\u002Ftroubleshooting.md) — failure classification and recovery sequence.\n- [Provenance](references\u002Frepo-provenance.md) — source revision and inspection evidence.\n- [Routing metadata](references\u002Frepo-routing-metadata.json) — scenario routing for the managed repo-skills router.\n\n## Completion checklist\n\nBefore calling a task complete, report:\n\n- the selected sub-skill and command\u002Fconfig used;\n- environment and backend assumptions (CPU, CUDA, Docker, external data, or credentials);\n- the output\u002Fevidence directory and the primary metric or artifact;\n- any warnings, blocked prerequisites, or unverified capabilities;\n- whether the run was a smoke check, a partial experiment, or a full result.\n","# RD-Agent\n\nUse this skill when a task involves the **RD-Agent** Python package or a checkout of the `microsoft\u002FRD-Agent` repository. RD-Agent is an agentic research framework, not a single model: it composes LLM calls, code generation, execution, evaluation, and experiment logs around several scenario families.\n\n## Route first\n\nChoose the smallest sub-skill that covers the user's requested workflow:\n\n| Request shape | Read next |\n|---|---|\n| Install, environment checks, CLI help, UI\u002Fserver startup, logs, config discovery, or safe maintenance | [setup-and-ops](sub-skills\u002Fsetup-and-ops\u002FSKILL.md) |\n| Qlib factors, factor reports, quantitative research, backtests, or finance data-agent\u002Fmodel-agent flows | [quant-finance](sub-skills\u002Fquant-finance\u002FSKILL.md) |\n| Kaggle-style tabular work, data-science agent loops, dataset preparation, or competition artifacts | [competition-data-science](sub-skills\u002Fcompetition-data-science\u002FSKILL.md) |\n| LLM fine-tuning datasets, training jobs, merge\u002Fevaluation, or fine-tune UI | [llm-finetune](sub-skills\u002Fllm-finetune\u002FSKILL.md) |\n| AutoRL-Bench, RL scenario loops, benchmark agents, or post-training experiments | [rl-post-training](sub-skills\u002Frl-post-training\u002FSKILL.md) |\n| General model copilot, paper-driven model exploration, or reusable model-research scaffolding | [paper-model-copilot](sub-skills\u002Fpaper-model-copilot\u002FSKILL.md) |\n\nFor a mixed request, read the relevant sub-skills in the order **setup → domain workflow → evaluation\u002Ftroubleshooting**. Do not load every sub-skill by default.\n\n## Operating model\n\n1. **Establish the runtime.** Confirm Python is supported (the inspected project supports Python 3.10 and 3.11), the package is importable, and the requested command is available. For a checkout, install with `python -m pip install -e .`; for an installed distribution, use the package's normal environment. Keep API keys, data roots, model caches, and experiment outputs outside the skill tree.\n2. **Select the scenario.** RD-Agent's public CLI exposes `data_science`, `fin_quant`, `fin_factor_report`, `llm_finetune`, `general_model`, `ui`, `server_ui`, and `ds_user_interact`. The RL benchmark entry point is `python -m rdagent.scenarios.rl.autorl_bench.run`.\n3. **Make configuration explicit.** Prefer an existing example or scenario config, then override only the values needed for the task. Record dataset paths, model\u002Fprovider settings, execution sandbox, output directory, and evaluation metric before launching an agent loop.\n4. **Run the smallest smoke check.** Use `--help`, import checks, or a tiny fixture before a long experiment. Do not infer a successful experiment from process startup: inspect the generated summary, evaluator result, and logs.\n5. **Preserve evidence.** Save the command, resolved configuration, commit\u002Fpackage version, metric definitions, and failure traceback with the experiment. Keep generated code and datasets in a separate run directory.\n\n## Common preflight\n\n```bash\npython -c \"import rdagent; print(rdagent.__file__)\"\nrdagent --help\nrdagent health_check --no-check-env --no-check-docker\n```\n\nThe health check can inspect Docker and external tools when those checks are enabled. Use the explicit `--no-check-*` flags only for a lightweight package\u002FCLI smoke test; do not treat that result as proof that a containerized or GPU-backed workflow is ready.\n\n## Safety and scope boundaries\n\n- Agent loops can generate and execute code. Start with a disposable output directory and a restricted execution environment.\n- Finance, Kaggle, model-training, and RL scenarios may require credentials, large datasets, Docker, CUDA, or external services. State those prerequisites instead of silently substituting a CPU or toy result.\n- Avoid launching full training, benchmark servers, UI processes, or network downloads as a generic smoke test. Use the bundled helper scripts for deterministic checks.\n- `fitz` may emit a deprecation warning: use of the warning alone is not a failure. AutoRL-Bench may report an empty registry when `SMITH_BENCH_DIR` is unset or points to a missing benchmark checkout.\n- Never place provider keys, cookies, competition credentials, private data, or local absolute paths in a generated report or a reusable configuration.\n\n## Evidence map\n\n- [Package and CLI overview](references\u002Fpackage-overview.md) — public entry points and component boundaries.\n- [Troubleshooting](references\u002Ftroubleshooting.md) — failure classification and recovery sequence.\n- [Provenance](references\u002Frepo-provenance.md) — source revision and inspection evidence.\n- [Routing metadata](references\u002Frepo-routing-metadata.json) — scenario routing for the managed repo-skills router.\n\n## Completion checklist\n\nBefore calling a task complete, report:\n\n- the selected sub-skill and command\u002Fconfig used;\n- environment and backend assumptions (CPU, CUDA, Docker, external data, or credentials);\n- the output\u002Fevidence directory and the primary metric or artifact;\n- any warnings, blocked prerequisites, or unverified capabilities;\n- whether the run was a smoke check, a partial experiment, or a full result.\n",{"name":5,"license":15,"metadata":24,"description":7,"disable-model-invocation":29},{"package":25,"disco-role":26,"source-revision":27,"source-repository":28},"rdagent","operating","6762f84f9bc0f5c6486c50a00e128a57ac6c3683","microsoft\u002FRD-Agent",true,"main","d69096e239f2a89aabc2cd078c47664a9d3b1e11","05-技能与代理开发\u002Frd-agent",3,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts","https:\u002F\u002Fgithub.com\u002Flza6","个人vx：Tf00798\r\n个人✈：@TFGY999","2026-10-08 14:29:22",{"source":8,"skillName":5,"commands":39,"prompt":49,"zipAvailable":29},[40,43,46],{"tool":41,"command":42},"npx","npx skills add lza6\u002FArmor-piercing-system-prompts --skill rd-agent",{"tool":44,"command":45},"bunx","bunx skills add lza6\u002FArmor-piercing-system-prompts --skill rd-agent",{"tool":47,"command":48},"pnpm","pnpm dlx skills add lza6\u002FArmor-piercing-system-prompts --skill rd-agent","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F05-技能与代理开发\u002Frd-agent\u002FSKILL.md",[51,57,61,65],{"path":52,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Fpackage-overview.md","REFERENCE","参考文档",2836,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F05-技能与代理开发\u002Frd-agent\u002Freferences\u002Fpackage-overview.md",{"path":58,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Frepo-provenance.md",2778,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F05-技能与代理开发\u002Frd-agent\u002Freferences\u002Frepo-provenance.md",{"path":62,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":63,"sourceHtmlUrl":64},"references\u002Frepo-routing-metadata.json",523,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F05-技能与代理开发\u002Frd-agent\u002Freferences\u002Frepo-routing-metadata.json",{"path":66,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":67,"sourceHtmlUrl":68},"references\u002Ftroubleshooting.md",3456,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F05-技能与代理开发\u002Frd-agent\u002Freferences\u002Ftroubleshooting.md",[70,77,82,89,96,103],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M7FFG485KY83TD6HSC4YPCMC","obsidian-clipper-template-creator","Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator\u002FSKILL.md",[],[],{"skillNo":78,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":19},"01M7FFG0T6AZ0STEJPSB3CX84K","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator-antigravity-awesome-skills-main\u002FSKILL.md",[],[],{"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":88,"publishedAt":19},"01M7FFFY0QEM14SKWPDYGMXM2P","python-api","Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP\u002FmAP semantics, and source-style import diagnostics.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Fpython-api\u002FSKILL.md",[],[],{"skillNo":90,"name":91,"nameZh":6,"description":92,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":19},"01M7FFFVZQZ2GJ0C0TGP7JHD20","file-evaluation","Folder-based PASCAL VOC AP\u002FmAP evaluation from ground-truth and detection text files.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Ffile-evaluation\u002FSKILL.md",[],[],{"skillNo":97,"name":98,"nameZh":6,"description":99,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":100,"categories":101,"keywords":102,"publishedAt":19},"01M7FFFSFNYSE77SJ6PE1CK2DA","object-detection-metrics","Guides legacy Object-Detection-Metrics PASCAL VOC AP\u002FmAP evaluation from detection text folders and direct Python API classes.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002FSKILL.md",[],[],{"skillNo":104,"name":105,"nameZh":6,"description":106,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":107,"categories":108,"keywords":109,"publishedAt":19},"01M7FFFP0GCK1RVXBCPQ8APBMT","nx-workspace-patterns","Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fnx-workspace-patterns\u002FSKILL.md",[],[]]