[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7FC59SSAYX8CGA01R4P6Z0B":3,"skill-same:01M7FC59SSAYX8CGA01R4P6Z0B":67},{"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":28,"commitSha":29,"skillDirectory":30,"forksCount":31,"repositoryHtmlUrl":32,"repositoryHomepageUrl":6,"ownerHtmlUrl":33,"ownerBlogUrl":6,"ownerDescription":34,"updatedAt":35,"install":36,"resources":48},"01M7FC59SSAYX8CGA01R4P6Z0B","baselines-and-projects",null,"Choose and safely operate Scenic baselines and project-specific research stacks, configs, registries, optional dependencies, checkpoints, datasets, and repo-owned tools.","lza6\u002FArmor-piercing-system-prompts","lza6","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F128880206?v=4","01M7EB3K7WPX5T28EJ2NVHE6EG","Chinese-tingfeng","01M7EB3K97E1KQ16QWKJMFNGX8",11,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fscenic\u002Fsub-skills\u002Fbaselines-and-projects\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: baselines-and-projects\ndescription: \"Choose and safely operate Scenic baselines and project-specific\n  research stacks, configs, registries, optional dependencies, checkpoints,\n  datasets, and repo-owned tools.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: Apache 2.0\n---\n\n# Baselines and Projects\n\nUse this sub-skill when a task is about **which Scenic project or baseline to start from**, how project-specific binaries\u002Fconfigs\u002Fregistries fit together, which optional dependency stack is needed, or whether a repo-owned conversion\u002Fevaluation tool is safe to run.\n\nDo **not** use this sub-skill for generic launch mechanics, low-level dataset registry implementation, or model\u002Flayer APIs:\n\n- Route `python -m ... --config=... --workdir=...`, config override syntax, workdir\u002Fcheckpoint hygiene, device flags, and restart\u002Feval-only mechanics to `running-and-training` after this sub-skill identifies the project and required config fields.\n- Route TFDS\u002FDMVR\u002FFlexIO dataset registration and preprocessing operator internals to `data-pipelines`.\n- Route Flax module signatures, model\u002Flayer implementation details, parameter trees, and checkpoint surgery to `modeling-and-layers`.\n\n## Fast decision procedure\n\n1. Identify the user's task family and modality.\n   - Video classification: start with ViViT for the canonical pure-transformer baseline, MTV for multi-view video recognition, TokenLearner for token-reduction experiments, AV-MAE for audio\u002Fvideo self-supervised pretraining or finetuning, and ObjectViViT when external object boxes are part of the method.\n   - Open-vocabulary or text-conditioned object detection: start with OWL-ViT. Use CLIP only for image\u002Ftext embedding baselines, not detection by itself.\n   - Dense video captioning: use Vid2Seq for offline single-stage dense captioning, Streaming DVC for long\u002Flow-latency streaming captions, and DenseVOC when the output must track and caption object trajectories.\n   - Generic image classification: use the baseline ViT, MLP-Mixer, ResNet\u002FBiT, Axial-ResNet, TokenLearner, MatViT, or PlainViT families depending on the architecture being compared.\n   - Detection\u002Fsegmentation baselines: use DETR, Deformable DETR, CenterNet\u002FCenterNet2, OWL-ViT, or SAM according to the supervision and prompt\u002Fquery style.\n2. Read [references\u002Fproject-catalog.md](references\u002Fproject-catalog.md) for the project family catalog, route selection table, project structure pattern, and project-specific main\u002Fconfig\u002Fregistry conventions.\n3. Read [references\u002Fbaseline-recipes.md](references\u002Fbaseline-recipes.md) when the user asks for a baseline recipe, a safe baseline config adaptation, or the difference between ViT\u002FResNet\u002FDETR\u002FCLIP\u002FBERT\u002FMLP-Mixer\u002FUNet\u002FCenterNet\u002FSAM families.\n4. Read [references\u002Foptional-dependencies.md](references\u002Foptional-dependencies.md) before installing project extras or interpreting missing imports. Install only the group required by the selected project.\n5. Read [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md) before running project-owned conversion\u002Fevaluation tools, when data\u002Fcheckpoints\u002Ftokenizers\u002FJava are missing, or when a custom registry reports an unsupported model\u002Ftrainer\u002Fdataset.\n6. Optionally run [scripts\u002Fproject_config_index.py](scripts\u002Fproject_config_index.py) against a user-supplied Scenic checkout to get a non-importing inventory of project config files, requirements files, and tool-script names before choosing a route.\n\n## Required context to collect before acting\n\nAsk or infer the following before recommending a project, installing extras, or approving a project tool:\n\n- Task output: classification labels, boxes, masks, temporal segments, dense captions, object trajectories, layout labels, point-cloud classes\u002Fparts, genomic variant classes, etc.\n- Modality: image, video, audio, audio+video, text+image, text+video, point cloud, UI layout, genomic pileup image.\n- Desired baseline type: stable baseline, newest project, reproduction target, transfer\u002Ffine-tune, inference only, or data conversion only.\n- Dataset status: already in TFDS\u002FDMVR\u002FFlexIO\u002FTFRecord form, raw files only, remote\u002Fprivate storage, or not yet acquired.\n- Checkpoint\u002Ftokenizer status: Scenic checkpoint, big_vision\u002FViT checkpoint, CLIP\u002FTorch checkpoint, T5\u002FT5X checkpoint, BERT vocab, SAM\u002FMAE\u002FVitDet conversion, or absent.\n- Compute and dependency budget: CPU smoke only, single GPU, multi-GPU\u002FTPU, strict pinned JAX\u002FFlax environment, or no network installs.\n\n## Safety rules\n\n- Treat project-owned `tools\u002F` scripts and conversion notebooks as **reference-only until prerequisites are proven**. Many tools read large external datasets, download or convert checkpoints, require Java or dataset credentials, and write TFRecords\u002FJSON\u002Fcheckpoints. See [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md#project-owned-tool-guardrails).\n- Do not run a conversion\u002Fevaluation tool merely because the user names it. First confirm all required input files, output location, expected runtime\u002Fstorage, credentials, and overwrite behavior.\n- If a requested project requires unavailable data, checkpoints, tokenizer files, Java, or accelerator memory, stop with a prerequisites list instead of silently switching projects.\n- If two projects require conflicting pins, create isolated environments; do not mix Deformable DETR's old JAX\u002FCUDA pins, Vid2Seq's Flax 0.5 expectation, Lingvo, and modern JAX extras in one environment.\n- Record any narrowed recommendation explicitly: e.g. \"OWL-ViT is the open-vocabulary detector, but with no LVIS\u002FCOCO annotations available I can only prepare an inference or config-inspection path.\"\n\n## Quick handoff templates\n\nProject selection handoff to `running-and-training`:\n\n```text\nSelected Scenic project: \u003Cproject family>\nReason: \u003Ctask\u002Fmodality\u002Fdesired output>\nEntrypoint module: \u003Cpython -m module name>\nConfig identifier: \u003Cpackage-relative config path string or config module>\nRequired config fields to set: \u003Cdataset paths, weights\u002Fcheckpoint, tokenizer, eval_only, batch size>\nOptional dependency group: \u003Cgroup from optional-dependencies.md>\nBlocked prerequisites: \u003Cnone or explicit data\u002Fcheckpoint\u002Fcompute\u002Fcredential list>\n```\n\nUnsafe tool response template:\n\n```text\nDo not run the project tool yet. It is a conversion\u002Fevaluation helper with external data\u002Fwrite side effects.\nPrerequisites to verify first: \u003Cinput dataset\u002Fannotation\u002Fcheckpoint\u002Ftokenizer>, \u003Coutput path>, \u003Cstorage\u002Fruntime>, \u003Ccredentials>, \u003Coverwrite policy>.\nSafe alternative now: use scripts\u002Fproject_config_index.py for static inventory, or prepare a dry prerequisite checklist.\n```\n","# Baselines and Projects\n\nUse this sub-skill when a task is about **which Scenic project or baseline to start from**, how project-specific binaries\u002Fconfigs\u002Fregistries fit together, which optional dependency stack is needed, or whether a repo-owned conversion\u002Fevaluation tool is safe to run.\n\nDo **not** use this sub-skill for generic launch mechanics, low-level dataset registry implementation, or model\u002Flayer APIs:\n\n- Route `python -m ... --config=... --workdir=...`, config override syntax, workdir\u002Fcheckpoint hygiene, device flags, and restart\u002Feval-only mechanics to `running-and-training` after this sub-skill identifies the project and required config fields.\n- Route TFDS\u002FDMVR\u002FFlexIO dataset registration and preprocessing operator internals to `data-pipelines`.\n- Route Flax module signatures, model\u002Flayer implementation details, parameter trees, and checkpoint surgery to `modeling-and-layers`.\n\n## Fast decision procedure\n\n1. Identify the user's task family and modality.\n   - Video classification: start with ViViT for the canonical pure-transformer baseline, MTV for multi-view video recognition, TokenLearner for token-reduction experiments, AV-MAE for audio\u002Fvideo self-supervised pretraining or finetuning, and ObjectViViT when external object boxes are part of the method.\n   - Open-vocabulary or text-conditioned object detection: start with OWL-ViT. Use CLIP only for image\u002Ftext embedding baselines, not detection by itself.\n   - Dense video captioning: use Vid2Seq for offline single-stage dense captioning, Streaming DVC for long\u002Flow-latency streaming captions, and DenseVOC when the output must track and caption object trajectories.\n   - Generic image classification: use the baseline ViT, MLP-Mixer, ResNet\u002FBiT, Axial-ResNet, TokenLearner, MatViT, or PlainViT families depending on the architecture being compared.\n   - Detection\u002Fsegmentation baselines: use DETR, Deformable DETR, CenterNet\u002FCenterNet2, OWL-ViT, or SAM according to the supervision and prompt\u002Fquery style.\n2. Read [references\u002Fproject-catalog.md](references\u002Fproject-catalog.md) for the project family catalog, route selection table, project structure pattern, and project-specific main\u002Fconfig\u002Fregistry conventions.\n3. Read [references\u002Fbaseline-recipes.md](references\u002Fbaseline-recipes.md) when the user asks for a baseline recipe, a safe baseline config adaptation, or the difference between ViT\u002FResNet\u002FDETR\u002FCLIP\u002FBERT\u002FMLP-Mixer\u002FUNet\u002FCenterNet\u002FSAM families.\n4. Read [references\u002Foptional-dependencies.md](references\u002Foptional-dependencies.md) before installing project extras or interpreting missing imports. Install only the group required by the selected project.\n5. Read [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md) before running project-owned conversion\u002Fevaluation tools, when data\u002Fcheckpoints\u002Ftokenizers\u002FJava are missing, or when a custom registry reports an unsupported model\u002Ftrainer\u002Fdataset.\n6. Optionally run [scripts\u002Fproject_config_index.py](scripts\u002Fproject_config_index.py) against a user-supplied Scenic checkout to get a non-importing inventory of project config files, requirements files, and tool-script names before choosing a route.\n\n## Required context to collect before acting\n\nAsk or infer the following before recommending a project, installing extras, or approving a project tool:\n\n- Task output: classification labels, boxes, masks, temporal segments, dense captions, object trajectories, layout labels, point-cloud classes\u002Fparts, genomic variant classes, etc.\n- Modality: image, video, audio, audio+video, text+image, text+video, point cloud, UI layout, genomic pileup image.\n- Desired baseline type: stable baseline, newest project, reproduction target, transfer\u002Ffine-tune, inference only, or data conversion only.\n- Dataset status: already in TFDS\u002FDMVR\u002FFlexIO\u002FTFRecord form, raw files only, remote\u002Fprivate storage, or not yet acquired.\n- Checkpoint\u002Ftokenizer status: Scenic checkpoint, big_vision\u002FViT checkpoint, CLIP\u002FTorch checkpoint, T5\u002FT5X checkpoint, BERT vocab, SAM\u002FMAE\u002FVitDet conversion, or absent.\n- Compute and dependency budget: CPU smoke only, single GPU, multi-GPU\u002FTPU, strict pinned JAX\u002FFlax environment, or no network installs.\n\n## Safety rules\n\n- Treat project-owned `tools\u002F` scripts and conversion notebooks as **reference-only until prerequisites are proven**. Many tools read large external datasets, download or convert checkpoints, require Java or dataset credentials, and write TFRecords\u002FJSON\u002Fcheckpoints. See [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md#project-owned-tool-guardrails).\n- Do not run a conversion\u002Fevaluation tool merely because the user names it. First confirm all required input files, output location, expected runtime\u002Fstorage, credentials, and overwrite behavior.\n- If a requested project requires unavailable data, checkpoints, tokenizer files, Java, or accelerator memory, stop with a prerequisites list instead of silently switching projects.\n- If two projects require conflicting pins, create isolated environments; do not mix Deformable DETR's old JAX\u002FCUDA pins, Vid2Seq's Flax 0.5 expectation, Lingvo, and modern JAX extras in one environment.\n- Record any narrowed recommendation explicitly: e.g. \"OWL-ViT is the open-vocabulary detector, but with no LVIS\u002FCOCO annotations available I can only prepare an inference or config-inspection path.\"\n\n## Quick handoff templates\n\nProject selection handoff to `running-and-training`:\n\n```text\nSelected Scenic project: \u003Cproject family>\nReason: \u003Ctask\u002Fmodality\u002Fdesired output>\nEntrypoint module: \u003Cpython -m module name>\nConfig identifier: \u003Cpackage-relative config path string or config module>\nRequired config fields to set: \u003Cdataset paths, weights\u002Fcheckpoint, tokenizer, eval_only, batch size>\nOptional dependency group: \u003Cgroup from optional-dependencies.md>\nBlocked prerequisites: \u003Cnone or explicit data\u002Fcheckpoint\u002Fcompute\u002Fcredential list>\n```\n\nUnsafe tool response template:\n\n```text\nDo not run the project tool yet. It is a conversion\u002Fevaluation helper with external data\u002Fwrite side effects.\nPrerequisites to verify first: \u003Cinput dataset\u002Fannotation\u002Fcheckpoint\u002Ftokenizer>, \u003Coutput path>, \u003Cstorage\u002Fruntime>, \u003Ccredentials>, \u003Coverwrite policy>.\nSafe alternative now: use scripts\u002Fproject_config_index.py for static inventory, or prepare a dry prerequisite checklist.\n```\n",{"name":5,"license":24,"metadata":25,"description":7,"disable-model-invocation":27},"Apache 2.0",{"disco-role":26},"operating",true,"main","d69096e239f2a89aabc2cd078c47664a9d3b1e11","06-前端与设计\u002Fscenic\u002Fsub-skills\u002Fbaselines-and-projects",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":37,"prompt":47,"zipAvailable":27},[38,41,44],{"tool":39,"command":40},"npx","npx skills add lza6\u002FArmor-piercing-system-prompts --skill baselines-and-projects",{"tool":42,"command":43},"bunx","bunx skills add lza6\u002FArmor-piercing-system-prompts --skill baselines-and-projects",{"tool":45,"command":46},"pnpm","pnpm dlx skills add lza6\u002FArmor-piercing-system-prompts --skill baselines-and-projects","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fscenic\u002Fsub-skills\u002Fbaselines-and-projects\u002FSKILL.md",[49,55,59,63],{"path":50,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":53,"sourceHtmlUrl":54},"references\u002Fbaseline-recipes.md","REFERENCE","参考文档",16041,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fscenic\u002Fsub-skills\u002Fbaselines-and-projects\u002Freferences\u002Fbaseline-recipes.md",{"path":56,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":57,"sourceHtmlUrl":58},"references\u002Foptional-dependencies.md",11019,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fscenic\u002Fsub-skills\u002Fbaselines-and-projects\u002Freferences\u002Foptional-dependencies.md",{"path":60,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":61,"sourceHtmlUrl":62},"references\u002Fproject-catalog.md",16646,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fscenic\u002Fsub-skills\u002Fbaselines-and-projects\u002Freferences\u002Fproject-catalog.md",{"path":64,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":65,"sourceHtmlUrl":66},"references\u002Ftroubleshooting.md",15083,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fscenic\u002Fsub-skills\u002Fbaselines-and-projects\u002Freferences\u002Ftroubleshooting.md",[68,76,81,88,95,102],{"skillNo":69,"name":70,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":72,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"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.","MIT","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator\u002FSKILL.md",[],[],{"skillNo":77,"name":70,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":72,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"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":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":72,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"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":89,"name":90,"nameZh":6,"description":91,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":72,"licenseFileUrl":6,"sourceHtmlUrl":92,"categories":93,"keywords":94,"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":96,"name":97,"nameZh":6,"description":98,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":72,"licenseFileUrl":6,"sourceHtmlUrl":99,"categories":100,"keywords":101,"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":103,"name":104,"nameZh":6,"description":105,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":72,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":108,"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",[],[]]