[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6DA2SNFM7QDR29R12HZGNQH":3,"skill-same:01M6DA2SNFM7QDR29R12HZGNQH":71},{"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":23,"publishedAt":34,"linkOnly":35,"markdownContent":36,"markdownBody":37,"frontMatter":38,"refName":39,"commitSha":40,"skillDirectory":41,"forksCount":42,"repositoryHtmlUrl":43,"repositoryHomepageUrl":6,"ownerHtmlUrl":44,"ownerBlogUrl":45,"ownerDescription":6,"updatedAt":46,"install":47,"resources":60},"01M6DA2SNFM7QDR29R12HZGNQH","scion-agent-manage",null,"Manage concurrent LLM-based code agents with scion - orchestrate parallel agents with isolated workspaces, troubleshoot and recover stuck agents","GoogleCloudPlatform\u002Fscion","GoogleCloudPlatform","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F2810941?v=4","01M6DA2FX07AS0KCQ4FBRS84A2","Google Cloud Platform","01M6DA2G1EA6A2E5MKG6THG0KK",2115,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fscion\u002Fblob\u002Fmain\u002Fresources\u002Fplatform_skills\u002Fscion-agent-manage\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[24,25,26,27,28,29,30,31,32,33],"manage","scion agent manage","troubleshoot","code","code agents","orchestrate","parallel agents","parallel","workspaces","with","2026-11-10 10:12:30",false,"---\nname: scion-agent-manage\ndescription: Manage concurrent LLM-based code agents with scion - orchestrate parallel agents with isolated workspaces, troubleshoot and recover stuck agents\n---\n\n# Scion Agent Management Skill\n\nScion is a container-based orchestration tool for managing concurrent LLM-based code agents. It enables parallel execution of specialized sub-agents with isolated identities, credentials, and workspaces.\n\n## Core Concepts\n\n### Projects\nA **project** is the grouping construct for agents in scion.\n\n### Agents\nAn **agent** is an isolated LLM instance running in a container with a mounted workspace, credentials, and configuration.\n\n### Templates\n**Templates** are blueprints for creating agents.\n\n### Harnesses\nA **harness** is the LLM interface (Gemini CLI, Claude Code, etc.) that the agent uses.\n\n## Command Reference\n\nThe best and most current reference for the CLI commands is available from `scion --help`. Some best practices are in the scion-cli-operations skill.\n\n## Tips for Agents\n\n1. **Check existing agents first**: Before starting a new agent, use `scion list` to see what's already running.\n\n2. **Use descriptive names**: Agent names should reflect their purpose (e.g., `refactor-auth`, `test-api`, `audit-security`).\n\n3. **Choose appropriate templates**: Use `--type researcher` for a researcher.\n\n4. **Monitor with logs**: Use `scion logs \u003Cagent>` to check progress without interrupting.\n\n5. **Interrupt carefully**: The `--interrupt` flag on messages stops current work - use only when necessary.\n\n6. **Preserve branches**: Use `--preserve-branch` to keep the branch after deletion for later review. The flag does not push — confirm the branch is on the remote first.\n\n7. **Agent state**: Do not attempt to resume an agent unless you were the one who stopped it. An 'idle' agent may still be working.\n\n## Creating Agents\n\nTo translate a natural-language request into a `scion start` command, map intent to flags:\n\n| Intent | Flag | Example |\n|---|---|---|\n| Agent role | `-t` \u002F `--type` | `-t developer`, `-t researcher`, `-t code-reviewer` |\n| LLM interface | `--harness` | `--harness claude`, `--harness gemini-cli` |\n| Model override | `--model` | `--model claude-sonnet-4-20250514` |\n\nExample — \"have a claude xl developer write a file\":\n\n```bash\nscion start file-writer -t developer --harness claude --model xl \\\n  \"Read your brief at \u002Fscion-volumes\u002Fscratchpad\u002Fbriefs\u002Ffile-writer.md and follow it.\"\n```\n\nAdditional options like `--thinking-level` (0–100) can tune agent reasoning depth. Run `scion start --help` for the full flag reference.\n\n## Briefing\n\nEvery agent you create needs a brief. Write the brief to a **shared scratchpad file and\npass the filepath** — do not inline a long brief into the creation command.\n\n```bash\nscion start \u003Cname> --non-interactive \\\n  \"Read your brief at \u002Fscion-volumes\u002Fscratchpad\u002Fbriefs\u002F\u003Cname>.md and follow it.\"\n```\n\nA brief states:\n\n| Section | Content |\n|---|---|\n| Task | what to do, in one or two sentences |\n| Context | what has already been decided, and where to read it |\n| Boundaries | what is explicitly out of scope |\n| Deliverable | what artifact is owed, and in what shape |\n| Reporting | who to report to, and when — including who to ask when blocked ([see below](#direct-questions-to-the-person-who-can-answer-them)) |\n\n### Direct questions to the person who can answer them\n\nWhen an agent needs a decision or input, it should ask the person named in the\nbrief's **Reporting** row — not relay through the coordinator unless the\ncoordinator *is* that person. An agent created to work with a specific user or\nlead already knows who to ask; routing the question through an intermediary\nwastes a round trip and risks the question being reframed in transit.\n\nWhen writing a brief, make the Reporting row explicit enough that the agent\nknows who to message for decisions. If different questions go to different\npeople, say so.\n\n### Waiting for the agent to start\n\nIn Hub mode `scion start` waits until the agent is running (or its launch\nfails) only when the Hub launches the agent asynchronously, after a workspace\nupload, or with `--attach`. Otherwise it returns as soon as the Hub answers, and\nthe agent may still be provisioning — check with `scion look \u003Cname>` or\n`scion list` before relying on it. When it waits, the default wait is the\nHub's remaining launch budget plus 30 seconds, or 5 minutes when the Hub does\nnot report one.\n\n- Pass `--wait-timeout \u003Cduration>` (for example `--wait-timeout 15m`) to wait\n  longer — needed if the Hub's launch timeout has been raised.\n- An exit status of 0 does not always mean the agent is running. When the\n  Hub accepted the launch but the wait could not follow it to running (the\n  wait timed out, or the agent's status is not readable with your\n  credential), `scion start` prints \"accepted by Hub and launching\" (with\n  `--format json`, the result carries a `launchNote` detail) and exits 0.\n  The launch keeps going and may still fail: check with `scion list` or\n  `scion look \u003Cname>` before relying on the agent. Do not retry the start\n  under a new name; that leaves a duplicate agent.\n- If the wait timed out, re-run `scion start \u003Cname>` to resume waiting; it\n  does not start a second agent.\n- Pass `--no-wait` to return as soon as the Hub accepts the launch; check\n  progress later with `scion look \u003Cname>` or `scion list`.\n- If the error says the create did not complete, the agent cannot be\n  restarted: delete it with `scion delete \u003Cname>` and create it again with\n  `scion start` using the same template and task (both are shown in the\n  error). If soft-delete retention is enabled on the Hub, the name stays\n  reserved until the agent is deleted with force=true or purged; until then,\n  use a new name.\n\nFor shell-escaping rules when passing prompts, see the `scion-cli-operations` skill —\ndo not improvise quoting.\n\n## Model Override\n\nTo start an agent with a specific model (overriding the harness default), use the `--model` flag:\n\n```bash\nscion start \u003Cname> --non-interactive --model medium\n```\n\n**Do NOT use `--harness-config` for this** — that flag expects a named harness configuration registered in the hub, not a model name.\n\nFor troubleshooting agents that are stalled, have hit an error, or are stuck see references\u002Ftroubleshooting.md\n\nFor agent lifecycle rules — when to delete, when to stop, who may authorize deletion, and how to reincarnate or move an agent to another broker — see references\u002Fagent-lifecycle.md\n","# Scion Agent Management Skill\n\nScion is a container-based orchestration tool for managing concurrent LLM-based code agents. It enables parallel execution of specialized sub-agents with isolated identities, credentials, and workspaces.\n\n## Core Concepts\n\n### Projects\nA **project** is the grouping construct for agents in scion.\n\n### Agents\nAn **agent** is an isolated LLM instance running in a container with a mounted workspace, credentials, and configuration.\n\n### Templates\n**Templates** are blueprints for creating agents.\n\n### Harnesses\nA **harness** is the LLM interface (Gemini CLI, Claude Code, etc.) that the agent uses.\n\n## Command Reference\n\nThe best and most current reference for the CLI commands is available from `scion --help`. Some best practices are in the scion-cli-operations skill.\n\n## Tips for Agents\n\n1. **Check existing agents first**: Before starting a new agent, use `scion list` to see what's already running.\n\n2. **Use descriptive names**: Agent names should reflect their purpose (e.g., `refactor-auth`, `test-api`, `audit-security`).\n\n3. **Choose appropriate templates**: Use `--type researcher` for a researcher.\n\n4. **Monitor with logs**: Use `scion logs \u003Cagent>` to check progress without interrupting.\n\n5. **Interrupt carefully**: The `--interrupt` flag on messages stops current work - use only when necessary.\n\n6. **Preserve branches**: Use `--preserve-branch` to keep the branch after deletion for later review. The flag does not push — confirm the branch is on the remote first.\n\n7. **Agent state**: Do not attempt to resume an agent unless you were the one who stopped it. An 'idle' agent may still be working.\n\n## Creating Agents\n\nTo translate a natural-language request into a `scion start` command, map intent to flags:\n\n| Intent | Flag | Example |\n|---|---|---|\n| Agent role | `-t` \u002F `--type` | `-t developer`, `-t researcher`, `-t code-reviewer` |\n| LLM interface | `--harness` | `--harness claude`, `--harness gemini-cli` |\n| Model override | `--model` | `--model claude-sonnet-4-20250514` |\n\nExample — \"have a claude xl developer write a file\":\n\n```bash\nscion start file-writer -t developer --harness claude --model xl \\\n  \"Read your brief at \u002Fscion-volumes\u002Fscratchpad\u002Fbriefs\u002Ffile-writer.md and follow it.\"\n```\n\nAdditional options like `--thinking-level` (0–100) can tune agent reasoning depth. Run `scion start --help` for the full flag reference.\n\n## Briefing\n\nEvery agent you create needs a brief. Write the brief to a **shared scratchpad file and\npass the filepath** — do not inline a long brief into the creation command.\n\n```bash\nscion start \u003Cname> --non-interactive \\\n  \"Read your brief at \u002Fscion-volumes\u002Fscratchpad\u002Fbriefs\u002F\u003Cname>.md and follow it.\"\n```\n\nA brief states:\n\n| Section | Content |\n|---|---|\n| Task | what to do, in one or two sentences |\n| Context | what has already been decided, and where to read it |\n| Boundaries | what is explicitly out of scope |\n| Deliverable | what artifact is owed, and in what shape |\n| Reporting | who to report to, and when — including who to ask when blocked ([see below](#direct-questions-to-the-person-who-can-answer-them)) |\n\n### Direct questions to the person who can answer them\n\nWhen an agent needs a decision or input, it should ask the person named in the\nbrief's **Reporting** row — not relay through the coordinator unless the\ncoordinator *is* that person. An agent created to work with a specific user or\nlead already knows who to ask; routing the question through an intermediary\nwastes a round trip and risks the question being reframed in transit.\n\nWhen writing a brief, make the Reporting row explicit enough that the agent\nknows who to message for decisions. If different questions go to different\npeople, say so.\n\n### Waiting for the agent to start\n\nIn Hub mode `scion start` waits until the agent is running (or its launch\nfails) only when the Hub launches the agent asynchronously, after a workspace\nupload, or with `--attach`. Otherwise it returns as soon as the Hub answers, and\nthe agent may still be provisioning — check with `scion look \u003Cname>` or\n`scion list` before relying on it. When it waits, the default wait is the\nHub's remaining launch budget plus 30 seconds, or 5 minutes when the Hub does\nnot report one.\n\n- Pass `--wait-timeout \u003Cduration>` (for example `--wait-timeout 15m`) to wait\n  longer — needed if the Hub's launch timeout has been raised.\n- An exit status of 0 does not always mean the agent is running. When the\n  Hub accepted the launch but the wait could not follow it to running (the\n  wait timed out, or the agent's status is not readable with your\n  credential), `scion start` prints \"accepted by Hub and launching\" (with\n  `--format json`, the result carries a `launchNote` detail) and exits 0.\n  The launch keeps going and may still fail: check with `scion list` or\n  `scion look \u003Cname>` before relying on the agent. Do not retry the start\n  under a new name; that leaves a duplicate agent.\n- If the wait timed out, re-run `scion start \u003Cname>` to resume waiting; it\n  does not start a second agent.\n- Pass `--no-wait` to return as soon as the Hub accepts the launch; check\n  progress later with `scion look \u003Cname>` or `scion list`.\n- If the error says the create did not complete, the agent cannot be\n  restarted: delete it with `scion delete \u003Cname>` and create it again with\n  `scion start` using the same template and task (both are shown in the\n  error). If soft-delete retention is enabled on the Hub, the name stays\n  reserved until the agent is deleted with force=true or purged; until then,\n  use a new name.\n\nFor shell-escaping rules when passing prompts, see the `scion-cli-operations` skill —\ndo not improvise quoting.\n\n## Model Override\n\nTo start an agent with a specific model (overriding the harness default), use the `--model` flag:\n\n```bash\nscion start \u003Cname> --non-interactive --model medium\n```\n\n**Do NOT use `--harness-config` for this** — that flag expects a named harness configuration registered in the hub, not a model name.\n\nFor troubleshooting agents that are stalled, have hit an error, or are stuck see references\u002Ftroubleshooting.md\n\nFor agent lifecycle rules — when to delete, when to stop, who may authorize deletion, and how to reincarnate or move an agent to another broker — see references\u002Fagent-lifecycle.md\n",{"name":5,"description":7},"main","d2283cad8aa80efb245b118b91c4accb684e287e","resources\u002Fplatform_skills\u002Fscion-agent-manage",310,"https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fscion","https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform","https:\u002F\u002Fcloud.google.com","2026-11-14 00:23:34",{"source":8,"skillName":5,"commands":48,"prompt":58,"zipAvailable":59},[49,52,55],{"tool":50,"command":51},"npx","npx skills add GoogleCloudPlatform\u002Fscion --skill scion-agent-manage",{"tool":53,"command":54},"bunx","bunx skills add GoogleCloudPlatform\u002Fscion --skill scion-agent-manage",{"tool":56,"command":57},"pnpm","pnpm dlx skills add GoogleCloudPlatform\u002Fscion --skill scion-agent-manage","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fscion\u002Fblob\u002Fmain\u002Fresources\u002Fplatform_skills\u002Fscion-agent-manage\u002FSKILL.md",true,[61,67],{"path":62,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":65,"sourceHtmlUrl":66},"references\u002Fagent-lifecycle.md","REFERENCE","参考文档",8431,"https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fscion\u002Fblob\u002Fmain\u002Fresources\u002Fplatform_skills\u002Fscion-agent-manage\u002Freferences\u002Fagent-lifecycle.md",{"path":68,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":69,"sourceHtmlUrl":70},"references\u002Ftroubleshooting.md",3319,"https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fscion\u002Fblob\u002Fmain\u002Fresources\u002Fplatform_skills\u002Fscion-agent-manage\u002Freferences\u002Ftroubleshooting.md",[72,106,140,164,181,192],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":79,"publishedAt":34},"01M6DA3C2RTK8R5TQ9YZVBYT1D","team-creation","Create or extend scion agent team templates from a high-level description of roles and workflow. Use when the user describes a multi-agent team, panel, crew, pipeline, or any scenario requiring coordinated LLM agents with distinct roles. Also use when adding new roles to an existing team, modifying workflows, or restructuring agent coordination. Produces ready-to-use template directories in .scion\u002Ftemplates\u002F by default, or in a custom path if specified.","https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fscion\u002Fblob\u002Fmain\u002Fresources\u002Fplatform_skills\u002Fteam-creation\u002FSKILL.md",[78],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105],"creation","team","team creation","ready","pipeline","default","use","llm agents","workflows","workflow","coordination","roles","create","agent team","template","crew","templates","scenario","multi-agent","user","directories","agent coordination","new","when","path","extend",{"skillNo":107,"name":108,"nameZh":6,"description":109,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":110,"categories":111,"keywords":112,"publishedAt":34},"01M6DA385P0S838W8ZXQ7SVDFM","scion-web-serving","How to serve a web page, app, API or other HTTP service from inside your agent so a person or program can reach it through the Hub's port proxy: scion expose and unexpose, the URL to hand over, what the proxy carries, and scion tunnel for full apps. The link serves web protocols only; for a non-HTTP port, point the person to scion tunnel. Use before running scion expose. For web pages and apps, also read references\u002Fweb-apps.md.","https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fscion\u002Fblob\u002Fmain\u002Fresources\u002Fplatform_skills\u002Fscion-web-serving\u002FSKILL.md",[],[113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,86,129,130,131,132,133,134,135,136,137,138,139],"scion web serving","web","serving","app","references","what the","protocols","how","proxy","page","hub","read","pages","tunnel","only","service","http","full","api","running","point","for","how to","apps","URL","program","hand",{"skillNo":141,"name":142,"nameZh":6,"description":143,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":144,"categories":145,"keywords":146,"publishedAt":34},"01M6DA35JGKREZP9ZM2ZYH0J7M","scion-scheduler","Schedule one-shot and recurring events for agents using the Scion CLI. Covers when to use scheduled events vs inline message delays, recurring cron schedules, lifecycle management, and the blocked-wait pairing rule.","https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fscion\u002Fblob\u002Fmain\u002Fresources\u002Fplatform_skills\u002Fscion-scheduler\u002FSKILL.md",[],[147,148,149,150,151,152,86,153,154,155,134,156,157,103,158,159,160,161,162,163],"scheduler","scion scheduler","recurring","message","cli","lifecycle","management","schedule","cron","schedules","scheduled","inline","when to use","rule","wait","pairing","events",{"skillNo":165,"name":166,"nameZh":6,"description":167,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":168,"categories":169,"keywords":170,"publishedAt":34},"01M6DA328HMWYY0R65MVVJJG3S","scion-messaging","Use when sending messages to users or agents with scion message, choosing conversations and mentions, or interpreting incoming messages. Covers user-facing formatting, agent delivery, and reading conversation history.","https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fscion\u002Fblob\u002Fmain\u002Fresources\u002Fplatform_skills\u002Fscion-messaging\u002FSKILL.md",[],[171,172,86,150,173,174,175,176,177,178,33,179,180,99,103],"messaging","scion messaging","conversation","delivery","history","conversations","messages","users","reading","formatting",{"skillNo":182,"name":183,"nameZh":6,"description":184,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":185,"categories":186,"keywords":187,"publishedAt":34},"01M6DA2YXJ384EVWJX6RPHDZ9R","scion-conversation","How to use the scion conversation command for managing conversations. Covers listing, creating, viewing, participant management, and message retrieval. Complements the scion-messaging skill.","https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fscion\u002Fblob\u002Fmain\u002Fresources\u002Fplatform_skills\u002Fscion-conversation\u002FSKILL.md",[],[173,188,171,150,189,120,86,153,176,134,135,190,191],"scion conversation","command","listing","retrieval",{"skillNo":193,"name":194,"nameZh":6,"description":195,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":196,"categories":197,"keywords":198,"publishedAt":34},"01M6DA2WCR1XJK9A3NSV2VGP56","scion-cli-operations","Operational constraints for Scion agents running in containerized sandboxes. Covers prohibited commands and hub-only API access. Complements the scion CLI reference and messaging skills.","https:\u002F\u002Fgithub.com\u002FGoogleCloudPlatform\u002Fscion\u002Fblob\u002Fmain\u002Fresources\u002Fplatform_skills\u002Fscion-cli-operations\u002FSKILL.md",[],[151,199,200,171,123,201,127,202,203,131,204,132,134,205,206,207,208],"operations","scion cli operations","commands","reference","access","constraints","cli reference","sandboxes","operational constraints","containerized"]