[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EDWD374K9MN92BKESEC082":3,"skill-same:01M7EDWD374K9MN92BKESEC082":66},{"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":27,"commitSha":28,"skillDirectory":29,"forksCount":14,"repositoryHtmlUrl":30,"repositoryHomepageUrl":6,"ownerHtmlUrl":31,"ownerBlogUrl":32,"ownerDescription":6,"updatedAt":33,"install":34,"resources":47},"01M7EDWD374K9MN92BKESEC082","molclaw-protein-ligand-mmpbsa",null,"Execution-ready protein-ligand MM\u002FGB(PB)SA workflow with explicit MCP handoffs and optional analysis.","sherkevin\u002Fbenchmark","sherkevin","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F78121652?v=4","01M7EDKCMQC9GKR7A7J45YNK7F","Shervin","01M7EDKCNA0A6ZHHG5XJ3VK1E5",0,"MIT","https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark\u002Fblob\u002Fcodex\u002Fprogress-20260924\u002Fsrc\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-protein-ligand-mmpbsa\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: molclaw-protein-ligand-mmpbsa\ndescription: Execution-ready protein-ligand MM\u002FGB(PB)SA workflow with explicit MCP handoffs and optional analysis.\nlicense: MIT license\nmetadata:\n    skill-author: PJLab\n---\n\n# Protein-Ligand MM\u002FPBSA Workflow (Execution-Ready)\n\nNote: \n- Local files are not directly accessible by the server. Please upload them to the server using `molclaw-file-transfer` before execution. \n- For PDB file inputs, it is recommended to preprocess them using `molclaw-pdbfixer` before execution.\n- Please refer to skill `molclaw-scp-server` to complete tool invocation.\n\n## Canonical Toolchain\n1. `fix_pdb` → cleans the receptor and returns `output_file`.\n2. `prepare_complex` → consumes the repaired protein plus ligand to build the MD workspace and writes `output_dir`.\n3. `run_mmpbsa` → consumes `work_dir` to execute GB\u002FPB calculations and emits CSV summaries.\n4. `analyze_mmpbsa` (optional) → consumes `run_mmpbsa.output_dir` to derive plots\u002Freports from the MM\u002FPBSA results.\n\n## SCP Tool Names (must use)\n- `fix_pdb`\n- `prepare_complex`\n- `run_mmpbsa`\n- `analyze_mmpbsa`\n\n## Tool References\n- `fix_pdb`: [reference_fix_pdb.md](reference_fix_pdb.md)\n- `prepare_complex`: [reference_prepare_complex.md](reference_prepare_complex.md)\n- `run_mmpbsa`: [reference_run_mmpbsa.md](reference_run_mmpbsa.md)\n- `analyze_mmpbsa`: [reference_analyze_mmpbsa.md](reference_analyze_mmpbsa.md)\n\n## Data Handover Contract\n1. `fix_pdb.output_file` → `prepare_complex.protein` (never substitute with manual paths).\n2. `prepare_complex.output_dir` → `run_mmpbsa.work_dir` (required to contain em.gro, md.xtc, md.tpr, topol.top, ligand files).\n3. `run_mmpbsa.output_dir` → `analyze_mmpbsa.work_dir` when analysis is enabled.\n\n## Step Details\n### Step 1: `fix_pdb`\n- **Pre-flight checks**: ensure `input_path` exists and is a readable PDB; `dry_run` defaults to `False` so the file is written unless explicitly skipping.\n- **Inputs**: `input_path`, `add_hydrogens`, `remove_water`, `replace_nonstandard`, `add_missing_residues`, `dry_run`.\n- **Optional stage**: set `keep_chains` to restrict the repair when a subset of chains is needed.\n- **Success criteria**: `status == \"success\"`, `output_file` is populated, atom\u002Fresidue counts report >0.\n- **Fallback hints**: replay the `msg`, mention which residue\u002Fchain failed, or rerun with `dry_run=True` to diagnose before writing.\n\n### Step 2: `prepare_complex`\n- **Pre-flight checks**: confirm `protein` equals `fix_pdb.output_file`, `ligand` path exists, and any `pose` index is valid within the ligand file.\n- **Inputs**: `protein`, `ligand`, `pose`, `gpu_ids`, `full_md`, `nvt_time`, `npt_time`, `md_time`, `ph`.\n- **Optional stage**: short circuits such as `full_md=False` for quick workspace generation, but then downstream steps must adjust expectations.\n- **Success criteria**: `status == \"success\"`, `output_dir` is non-empty, `files` contains `em.gro`\u002F`md.xtc`\u002F`topol.top`.\n- **Fallback hints**: if GPU resources are oversubscribed, rerun with `gpu_ids` trimmed or `full_md=False`; if ligand conversion fails, switch to `.mol2` or `.pdb` input.\n\n### Step 3: `run_mmpbsa`\n- **Pre-flight checks**: `work_dir` exists and matches `prepare_complex.output_dir`; verify expected `em.gro`, `md.xtc`, and `topol.top` exist before invoking.\n- **Inputs**: `work_dir`, `method` (gb\u002Fpb\u002Fboth), `nproc`, `interval`, `startframe`, `endframe`, `generate_input`, `dry_run`.\n- **Optional stage**: `method` controls whether GB, PB, or both runs execute; `dry_run=True` tests configuration without executing heavy compute.\n- **Success criteria**: `status == \"success\"`, `gb_dir` or `pb_dir` is created, `results` dictionary contains energy estimates, required CSVs (`FINAL_RESULTS.csv`) exist.\n- **Fallback hints**: on missing CSVs, check `nproc` limits or re-run with `generate_input=True`; if `dry_run` was left `True`, rerun with the actual run.\n\n### Step 4: `analyze_mmpbsa` (optional)\n- **Pre-flight checks**: `enable_analysis` flag `True` and `work_dir` equals `run_mmpbsa.output_dir`; confirm the CSV outputs are present.\n- **Inputs**: `work_dir`.\n- **Optional stage**: no extra per-file overrides are required in normal workflow usage.\n- **Success criteria**: `status == \"success\"`, `output_dir` exists, `reports` dictionary lists generated PNG\u002FCSV\u002FMD artifacts.\n- **Fallback hints**: treat analyzer failures as soft; log the `msg` and keep the `run_mmpbsa` results as the primary output.\n\n## Agent Flow\n1. **Protein repair → complex build**: `fix_pdb.output_file` feeds `prepare_complex.protein`, guaranteeing the complex uses the cleaned receptor.\n2. **Workspace build → free-energy run**: `prepare_complex.output_dir` becomes `run_mmpbsa.work_dir`, producing both `mmgbsa\u002FFINAL_RESULTS.csv` and `mmpbsa\u002FFINAL_RESULTS.csv` depending on `method`.\n3. **Optional analysis**: when enabled, the analyzer consumes `run_mmpbsa.output_dir` (`work_dir`) to assemble summary plots, which are then surfaced through `analyze_mmpbsa.output_dir` and `reports`.\n4. **Normalized insights**: every branch funnels into the same `work_dir` CSV artifacts so downstream agents can compare GB, PB, and analyzer outputs uniformly.\n\n## Normalized Outputs (per workflow)\n- `fix_pdb`: `output_file` (cleaned receptor).\n- `prepare_complex`: `output_dir`\u002F`files` (workspace containing GROMACS artifacts).\n- `run_mmpbsa`: `work_dir`, `gb_dir`, `pb_dir`, `results`, `command`.\n- `analyze_mmpbsa`: `output_dir`, `reports`, `detected_mode`, `missing_files` (optional insights).\n\n## Safety Checklist\n- Never pass manual MD files instead of `prepare_complex.output_dir`.\n- Treat any missing `output_file`\u002F`output_dir` as a hard failure and stop before the next tool.\n- Validate GPU\u002FCPU resources before running long `run_mmpbsa` jobs (`nproc`, `gpu_ids`).\n- If `method=\"both\"`, ensure both PB and GB CSVs are created; otherwise, rerun with `method=\"pb\"` or `\"gb\"` only.\n- Surface actionable diagnostics (`missing file`, `invalid pose`, `transport error`) in the `msg` field.\n","# Protein-Ligand MM\u002FPBSA Workflow (Execution-Ready)\n\nNote: \n- Local files are not directly accessible by the server. Please upload them to the server using `molclaw-file-transfer` before execution. \n- For PDB file inputs, it is recommended to preprocess them using `molclaw-pdbfixer` before execution.\n- Please refer to skill `molclaw-scp-server` to complete tool invocation.\n\n## Canonical Toolchain\n1. `fix_pdb` → cleans the receptor and returns `output_file`.\n2. `prepare_complex` → consumes the repaired protein plus ligand to build the MD workspace and writes `output_dir`.\n3. `run_mmpbsa` → consumes `work_dir` to execute GB\u002FPB calculations and emits CSV summaries.\n4. `analyze_mmpbsa` (optional) → consumes `run_mmpbsa.output_dir` to derive plots\u002Freports from the MM\u002FPBSA results.\n\n## SCP Tool Names (must use)\n- `fix_pdb`\n- `prepare_complex`\n- `run_mmpbsa`\n- `analyze_mmpbsa`\n\n## Tool References\n- `fix_pdb`: [reference_fix_pdb.md](reference_fix_pdb.md)\n- `prepare_complex`: [reference_prepare_complex.md](reference_prepare_complex.md)\n- `run_mmpbsa`: [reference_run_mmpbsa.md](reference_run_mmpbsa.md)\n- `analyze_mmpbsa`: [reference_analyze_mmpbsa.md](reference_analyze_mmpbsa.md)\n\n## Data Handover Contract\n1. `fix_pdb.output_file` → `prepare_complex.protein` (never substitute with manual paths).\n2. `prepare_complex.output_dir` → `run_mmpbsa.work_dir` (required to contain em.gro, md.xtc, md.tpr, topol.top, ligand files).\n3. `run_mmpbsa.output_dir` → `analyze_mmpbsa.work_dir` when analysis is enabled.\n\n## Step Details\n### Step 1: `fix_pdb`\n- **Pre-flight checks**: ensure `input_path` exists and is a readable PDB; `dry_run` defaults to `False` so the file is written unless explicitly skipping.\n- **Inputs**: `input_path`, `add_hydrogens`, `remove_water`, `replace_nonstandard`, `add_missing_residues`, `dry_run`.\n- **Optional stage**: set `keep_chains` to restrict the repair when a subset of chains is needed.\n- **Success criteria**: `status == \"success\"`, `output_file` is populated, atom\u002Fresidue counts report >0.\n- **Fallback hints**: replay the `msg`, mention which residue\u002Fchain failed, or rerun with `dry_run=True` to diagnose before writing.\n\n### Step 2: `prepare_complex`\n- **Pre-flight checks**: confirm `protein` equals `fix_pdb.output_file`, `ligand` path exists, and any `pose` index is valid within the ligand file.\n- **Inputs**: `protein`, `ligand`, `pose`, `gpu_ids`, `full_md`, `nvt_time`, `npt_time`, `md_time`, `ph`.\n- **Optional stage**: short circuits such as `full_md=False` for quick workspace generation, but then downstream steps must adjust expectations.\n- **Success criteria**: `status == \"success\"`, `output_dir` is non-empty, `files` contains `em.gro`\u002F`md.xtc`\u002F`topol.top`.\n- **Fallback hints**: if GPU resources are oversubscribed, rerun with `gpu_ids` trimmed or `full_md=False`; if ligand conversion fails, switch to `.mol2` or `.pdb` input.\n\n### Step 3: `run_mmpbsa`\n- **Pre-flight checks**: `work_dir` exists and matches `prepare_complex.output_dir`; verify expected `em.gro`, `md.xtc`, and `topol.top` exist before invoking.\n- **Inputs**: `work_dir`, `method` (gb\u002Fpb\u002Fboth), `nproc`, `interval`, `startframe`, `endframe`, `generate_input`, `dry_run`.\n- **Optional stage**: `method` controls whether GB, PB, or both runs execute; `dry_run=True` tests configuration without executing heavy compute.\n- **Success criteria**: `status == \"success\"`, `gb_dir` or `pb_dir` is created, `results` dictionary contains energy estimates, required CSVs (`FINAL_RESULTS.csv`) exist.\n- **Fallback hints**: on missing CSVs, check `nproc` limits or re-run with `generate_input=True`; if `dry_run` was left `True`, rerun with the actual run.\n\n### Step 4: `analyze_mmpbsa` (optional)\n- **Pre-flight checks**: `enable_analysis` flag `True` and `work_dir` equals `run_mmpbsa.output_dir`; confirm the CSV outputs are present.\n- **Inputs**: `work_dir`.\n- **Optional stage**: no extra per-file overrides are required in normal workflow usage.\n- **Success criteria**: `status == \"success\"`, `output_dir` exists, `reports` dictionary lists generated PNG\u002FCSV\u002FMD artifacts.\n- **Fallback hints**: treat analyzer failures as soft; log the `msg` and keep the `run_mmpbsa` results as the primary output.\n\n## Agent Flow\n1. **Protein repair → complex build**: `fix_pdb.output_file` feeds `prepare_complex.protein`, guaranteeing the complex uses the cleaned receptor.\n2. **Workspace build → free-energy run**: `prepare_complex.output_dir` becomes `run_mmpbsa.work_dir`, producing both `mmgbsa\u002FFINAL_RESULTS.csv` and `mmpbsa\u002FFINAL_RESULTS.csv` depending on `method`.\n3. **Optional analysis**: when enabled, the analyzer consumes `run_mmpbsa.output_dir` (`work_dir`) to assemble summary plots, which are then surfaced through `analyze_mmpbsa.output_dir` and `reports`.\n4. **Normalized insights**: every branch funnels into the same `work_dir` CSV artifacts so downstream agents can compare GB, PB, and analyzer outputs uniformly.\n\n## Normalized Outputs (per workflow)\n- `fix_pdb`: `output_file` (cleaned receptor).\n- `prepare_complex`: `output_dir`\u002F`files` (workspace containing GROMACS artifacts).\n- `run_mmpbsa`: `work_dir`, `gb_dir`, `pb_dir`, `results`, `command`.\n- `analyze_mmpbsa`: `output_dir`, `reports`, `detected_mode`, `missing_files` (optional insights).\n\n## Safety Checklist\n- Never pass manual MD files instead of `prepare_complex.output_dir`.\n- Treat any missing `output_file`\u002F`output_dir` as a hard failure and stop before the next tool.\n- Validate GPU\u002FCPU resources before running long `run_mmpbsa` jobs (`nproc`, `gpu_ids`).\n- If `method=\"both\"`, ensure both PB and GB CSVs are created; otherwise, rerun with `method=\"pb\"` or `\"gb\"` only.\n- Surface actionable diagnostics (`missing file`, `invalid pose`, `transport error`) in the `msg` field.\n",{"name":5,"license":24,"metadata":25,"description":7},"MIT license",{"skill-author":26},"PJLab","codex\u002Fprogress-20260924","c8ee42aa1da4aadc7e9b7e39b01c27ccdeea5cc9","src\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-protein-ligand-mmpbsa","https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark","https:\u002F\u002Fgithub.com\u002Fsherkevin","https:\u002F\u002Fwww.cnblogs.com\u002Fsherkevin","2026-10-09 16:43:49",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add sherkevin\u002Fbenchmark --skill molclaw-protein-ligand-mmpbsa",{"tool":40,"command":41},"bunx","bunx skills add sherkevin\u002Fbenchmark --skill molclaw-protein-ligand-mmpbsa",{"tool":43,"command":44},"pnpm","pnpm dlx skills add sherkevin\u002Fbenchmark --skill molclaw-protein-ligand-mmpbsa","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark\u002Fblob\u002Fcodex\u002Fprogress-20260924\u002Fsrc\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-protein-ligand-mmpbsa\u002FSKILL.md",true,[48,54,58,62],{"path":49,"resourceType":50,"resourceTypeDesc":51,"sizeBytes":52,"sourceHtmlUrl":53},"reference_analyze_mmpbsa.md","REFERENCE","参考文档",1722,"https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark\u002Fblob\u002Fcodex\u002Fprogress-20260924\u002Fsrc\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-protein-ligand-mmpbsa\u002Freference_analyze_mmpbsa.md",{"path":55,"resourceType":50,"resourceTypeDesc":51,"sizeBytes":56,"sourceHtmlUrl":57},"reference_fix_pdb.md",2239,"https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark\u002Fblob\u002Fcodex\u002Fprogress-20260924\u002Fsrc\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-protein-ligand-mmpbsa\u002Freference_fix_pdb.md",{"path":59,"resourceType":50,"resourceTypeDesc":51,"sizeBytes":60,"sourceHtmlUrl":61},"reference_prepare_complex.md",2431,"https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark\u002Fblob\u002Fcodex\u002Fprogress-20260924\u002Fsrc\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-protein-ligand-mmpbsa\u002Freference_prepare_complex.md",{"path":63,"resourceType":50,"resourceTypeDesc":51,"sizeBytes":64,"sourceHtmlUrl":65},"reference_run_mmpbsa.md",2248,"https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark\u002Fblob\u002Fcodex\u002Fprogress-20260924\u002Fsrc\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-protein-ligand-mmpbsa\u002Freference_run_mmpbsa.md",[67,74,81,88,95,102],{"skillNo":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":19},"01M7EDY1139HR68WB7A3B7S4SC","molclaw-smiles-valid-check","Check if the input molecule SMILES string is valid.","https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark\u002Fblob\u002Fcodex\u002Fprogress-20260924\u002Fsrc\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-smiles-valid-check\u002FSKILL.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M7EDXNFE3A53X9RJHZ7MEVR1","molclaw-skill-template-writer","Formats extracted execution patterns into standard MolClaw skill documents. Accepts structured input from the Skill Crystallization Meta-Workflow (L2-12) and outputs a properly formatted L1 or L2 skill document conforming to MolClaw conventions. This skill ensures that auto-generated skills are structurally identical to expert-curated skills, enabling seamless integration into the skill matching and loading pipeline.","https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark\u002Fblob\u002Fcodex\u002Fprogress-20260924\u002Fsrc\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-skill-template-writer\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":15,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":19},"01M7EDXJSJESBRR1DQYRCS4DTW","molclaw-sequence-valid-check","Check if the input protein sequence is valid.","https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark\u002Fblob\u002Fcodex\u002Fprogress-20260924\u002Fsrc\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-sequence-valid-check\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":15,"licenseFileUrl":6,"sourceHtmlUrl":92,"categories":93,"keywords":94,"publishedAt":19},"01M7EDXF2M82DJKKFBTPV1FJFN","molclaw-scp-server","All tools utilized within molclaw skills connect via the MCP protocol. This skill serves as a unified guide for using the MCP Server. This skill must be loaded to create the MCP server before invoking any tools.","https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark\u002Fblob\u002Fcodex\u002Fprogress-20260924\u002Fsrc\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-scp-server\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":15,"licenseFileUrl":6,"sourceHtmlUrl":99,"categories":100,"keywords":101,"publishedAt":19},"01M7EDXBSW1SMA87ZMYJVGD1NW","molclaw-run-bioemu","Run BioEmu sequence sampling and extract ensemble structures for downstream conformation analysis.","https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark\u002Fblob\u002Fcodex\u002Fprogress-20260924\u002Fsrc\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-run-bioemu\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":15,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":108,"publishedAt":19},"01M7EDX8VMARAB5WTQ1DDVMY1P","molclaw-rgroup-sampling","Generate new molecules sampling from the input scaffold.","https:\u002F\u002Fgithub.com\u002Fsherkevin\u002Fbenchmark\u002Fblob\u002Fcodex\u002Fprogress-20260924\u002Fsrc\u002Fmolbench_repo\u002Fskills\u002FL1_tools\u002Fmolclaw-rgroup-sampling\u002FSKILL.md",[],[]]