[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M730050TZGYR7G5QR5AYSMYA":3,"skill-same:01M730050TZGYR7G5QR5AYSMYA":48},{"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":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":29,"ownerHtmlUrl":30,"ownerBlogUrl":31,"ownerDescription":32,"updatedAt":33,"install":34,"resources":47},"01M730050TZGYR7G5QR5AYSMYA","ghostlab",null,"Run a configured coding agent (Claude Code, Codex, Copilot CLI, OpenCode) headlessly on a task — locally or sandboxed in NVIDIA OpenShell — or write a Ghostlab pipeline that chains agent calls into a checkpointed, resumable job (evaluations such as MCP or skill evals, dataset labeling, test generation, any batch agent work). Use when asked to run an agent on a task in a sandbox, to run many agent tasks over a dataset, to build an agentic eval, or to resume or inspect a Ghostlab run.","sajjadGG\u002FGhostlab","sajjadGG","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F46228131?v=4","01M73000ZNS6D1HGN61GSHDVY0","Sajad","01M730010EJ7MFRXDMESPZ50W0",3,"MIT","https:\u002F\u002Fgithub.com\u002FsajjadGG\u002FGhostlab\u002Fblob\u002Fmain\u002Fskills\u002Fghostlab\u002FSKILL.md",[],[],"2026-11-11 15:11:14",false,"---\nname: ghostlab\ndescription: Run a configured coding agent (Claude Code, Codex, Copilot CLI, OpenCode) headlessly on a task — locally or sandboxed in NVIDIA OpenShell — or write a Ghostlab pipeline that chains agent calls into a checkpointed, resumable job (evaluations such as MCP or skill evals, dataset labeling, test generation, any batch agent work). Use when asked to run an agent on a task in a sandbox, to run many agent tasks over a dataset, to build an agentic eval, or to resume or inspect a Ghostlab run.\n---\n\n# Ghostlab\n\nGhostlab is infrastructure for headless agent work:\n\n- an **agent** (YAML\u002FJSON or Python) = harness + model + instructions + skills\n  + MCPs + workspace + sandbox;\n- a **run** of it = `Result` with `status`, `answer`, `tool_calls`, `.json()`;\n- a **pipeline** = stages that map, fan out, or gather items, where every item\n  and every agent call is checkpointed, so re-running resumes.\n\n## One task\n\n```bash\nghostlab exec --agent agent.yaml \"task\" [--sandbox local] [--collect OUT.md] [--json]\n```\n\nA minimal agent:\n\n```yaml\nharness: claude        # claude | codex | copilot | opencode | command\nmodel: sonnet\ninstructions: Be concise.\nskills: [skills\u002Fx]\nmcps: {docs: {command: node, args: [docs-mcp.js]}}\nworkspace: repo        # always a private copy\nsandbox: {backend: openshell, providers: [anthropic]}   # or: sandbox: local\n```\n\nThe answer goes to stdout. The trace is in `runs\u002F\u003Cstamp>-\u003Cname>\u002Fturn-1\u002F`\n(`prompt.txt`, `stdout.txt`, `stderr.txt`, `answer.md`, `result.json`).\n\n## Many tasks: write a pipeline\n\n```python\nfrom ghostlab import Agent, Pipeline, stage\n\n@stage(concurrency=4, retries=1)\ndef work(ctx, item):\n    reply = ctx.run(ctx.agent(\"worker\"), f\"... {item['text']} ... Reply with JSON.\")\n    data = reply.json()                 # raise if the answer is unusable -> retried\n    return {**item, **data}\n\npipeline = Pipeline(\"job\", [work], params={\"worker\": Agent.load(\"worker.yaml\")})\n```\n\n```bash\nghostlab run job.py --input items.jsonl         # results: runs\u002Fjob\u002Fresults.jsonl\nghostlab run job.py                             # resume after a crash or a failure\nghostlab status runs\u002Fjob\n```\n\nBuilding blocks:\n\n- `@stage(fan_out=True)` returns a list, and each element becomes an item;\n  `@stage(gather=True)` receives all items at once (for reports); returning\n  `None` from a map stage drops the item.\n- `ctx.session(agent)` holds a multi-turn conversation that resumes the\n  harness session — use it for emulated users.\n- `agent.with_(model=..., instructions=...)` and `agent.with_mcp(name, spec)`\n  derive variants. Pass agents as params so `--param worker=other.yaml` swaps\n  them without code changes.\n- Bump `@stage(version=\"2\")` after changing a stage's logic to recompute it.\n\n`examples\u002Fmcp-eval\u002Fpipeline.py` is a full evaluation pipeline: describe the\nMCP → generate scenarios → run emulated-user conversations and judge them →\nreport.\n\n## Reading failures\n\n`Result.status` separates the causes:\n\n| Status | Meaning — what to do |\n| --- | --- |\n| `agent_error` | the harness exited non-zero; read `turn-N\u002Fstderr.txt` |\n| `model_unavailable` | quota, auth, or an unsupported model; change the model or the credentials |\n| `timed_out` | raise the agent's `timeout` |\n| `sandbox_error` | OpenShell failed (gateway, image, policy); check `openshell status` and `openshell.log` |\n| `harness_error` | the CLI binary was not found |\n\nA failed pipeline item keeps its traceback in `\u003CNN-stage>\u002F\u003Ckey>\u002Fitem.json` and\nthe traces of failed attempts under `attempts\u002FN\u002F`.\n","# Ghostlab\n\nGhostlab is infrastructure for headless agent work:\n\n- an **agent** (YAML\u002FJSON or Python) = harness + model + instructions + skills\n  + MCPs + workspace + sandbox;\n- a **run** of it = `Result` with `status`, `answer`, `tool_calls`, `.json()`;\n- a **pipeline** = stages that map, fan out, or gather items, where every item\n  and every agent call is checkpointed, so re-running resumes.\n\n## One task\n\n```bash\nghostlab exec --agent agent.yaml \"task\" [--sandbox local] [--collect OUT.md] [--json]\n```\n\nA minimal agent:\n\n```yaml\nharness: claude        # claude | codex | copilot | opencode | command\nmodel: sonnet\ninstructions: Be concise.\nskills: [skills\u002Fx]\nmcps: {docs: {command: node, args: [docs-mcp.js]}}\nworkspace: repo        # always a private copy\nsandbox: {backend: openshell, providers: [anthropic]}   # or: sandbox: local\n```\n\nThe answer goes to stdout. The trace is in `runs\u002F\u003Cstamp>-\u003Cname>\u002Fturn-1\u002F`\n(`prompt.txt`, `stdout.txt`, `stderr.txt`, `answer.md`, `result.json`).\n\n## Many tasks: write a pipeline\n\n```python\nfrom ghostlab import Agent, Pipeline, stage\n\n@stage(concurrency=4, retries=1)\ndef work(ctx, item):\n    reply = ctx.run(ctx.agent(\"worker\"), f\"... {item['text']} ... Reply with JSON.\")\n    data = reply.json()                 # raise if the answer is unusable -> retried\n    return {**item, **data}\n\npipeline = Pipeline(\"job\", [work], params={\"worker\": Agent.load(\"worker.yaml\")})\n```\n\n```bash\nghostlab run job.py --input items.jsonl         # results: runs\u002Fjob\u002Fresults.jsonl\nghostlab run job.py                             # resume after a crash or a failure\nghostlab status runs\u002Fjob\n```\n\nBuilding blocks:\n\n- `@stage(fan_out=True)` returns a list, and each element becomes an item;\n  `@stage(gather=True)` receives all items at once (for reports); returning\n  `None` from a map stage drops the item.\n- `ctx.session(agent)` holds a multi-turn conversation that resumes the\n  harness session — use it for emulated users.\n- `agent.with_(model=..., instructions=...)` and `agent.with_mcp(name, spec)`\n  derive variants. Pass agents as params so `--param worker=other.yaml` swaps\n  them without code changes.\n- Bump `@stage(version=\"2\")` after changing a stage's logic to recompute it.\n\n`examples\u002Fmcp-eval\u002Fpipeline.py` is a full evaluation pipeline: describe the\nMCP → generate scenarios → run emulated-user conversations and judge them →\nreport.\n\n## Reading failures\n\n`Result.status` separates the causes:\n\n| Status | Meaning — what to do |\n| --- | --- |\n| `agent_error` | the harness exited non-zero; read `turn-N\u002Fstderr.txt` |\n| `model_unavailable` | quota, auth, or an unsupported model; change the model or the credentials |\n| `timed_out` | raise the agent's `timeout` |\n| `sandbox_error` | OpenShell failed (gateway, image, policy); check `openshell status` and `openshell.log` |\n| `harness_error` | the CLI binary was not found |\n\nA failed pipeline item keeps its traceback in `\u003CNN-stage>\u002F\u003Ckey>\u002Fitem.json` and\nthe traces of failed attempts under `attempts\u002FN\u002F`.\n",{"name":5,"description":7},"main","3c3a028fccc6a0e643a48b8e37b38136b2de8252","skills\u002Fghostlab",0,"https:\u002F\u002Fgithub.com\u002FsajjadGG\u002FGhostlab","https:\u002F\u002Fsajjadgg.github.io\u002FGhostlab\u002F","https:\u002F\u002Fgithub.com\u002FsajjadGG","https:\u002F\u002Fwww.linkedin.com\u002Fin\u002Fsajjadgg\u002F","Just a simple man","2026-11-07 13:22:05",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add sajjadGG\u002FGhostlab --skill ghostlab",{"tool":40,"command":41},"bunx","bunx skills add sajjadGG\u002FGhostlab --skill ghostlab",{"tool":43,"command":44},"pnpm","pnpm dlx skills add sajjadGG\u002FGhostlab --skill ghostlab","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FsajjadGG\u002FGhostlab\u002Fblob\u002Fmain\u002Fskills\u002Fghostlab\u002FSKILL.md",true,[],[49],{"skillNo":50,"name":51,"nameZh":6,"description":52,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":53,"categories":54,"keywords":55,"publishedAt":19},"01M73003J32J7TCWTRHPJ345NY","release-notes","Draft release notes from a changelog","https:\u002F\u002Fgithub.com\u002FsajjadGG\u002FGhostlab\u002Fblob\u002Fmain\u002Fexamples\u002Frelease-notes\u002Fskills\u002Frelease-notes\u002FSKILL.md",[],[]]