[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M760N1P2VTGVE6C4N5WF3FVR":3,"skill-same:01M760N1P2VTGVE6C4N5WF3FVR":47},{"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":6,"ownerHtmlUrl":29,"ownerBlogUrl":30,"ownerDescription":31,"updatedAt":32,"install":33,"resources":46},"01M760N1P2VTGVE6C4N5WF3FVR","workflow-reference",null,"Field reference for Workflow YAML (root keys, every state type, bindings, outputs, runtime error codes), what the engine guarantees, and a review fan-out example. For design, call load_skill with name workflow-authoring.","microsoft\u002FPilotSwarm","microsoft","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F6154722?v=4","01M698VYDXZ6JW5WH3CRV650R8","Microsoft","01M69A6RNK492A5JSNBTXANKX5",16,"MIT","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FPilotSwarm\u002Fblob\u002Fmain\u002Fpackages\u002Fsdk\u002Fplugins\u002Fworkflow\u002Fskills\u002Fworkflow-reference\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: workflow-reference\ndescription: Field reference for Workflow YAML (root keys, every state type, bindings, outputs, runtime error codes), what the engine guarantees, and a review fan-out example. For design, call load_skill with name workflow-authoring.\n---\n\n# Workflow reference\n\nExact fields for `apiVersion: pilotswarm.dev\u002Fv1` definitions and the\nworkflow tools. `workflow-authoring` explains how to design with them.\n\nIds: a state id or an outcome is a letter, then at most 63 letters, digits,\n`_` or `-`. State ids are unique in the whole workflow, body and branch\nstates included. A contract name is `Name@N` (for example `BuildEvent@1`).\n\n## Root keys\n\n| Key | Rule |\n|---|---|\n| `apiVersion` | `pilotswarm.dev\u002Fv1` |\n| `kind` | `Workflow` |\n| `metadata` | `{ name, version, description? }`. `name`: at most 128 characters. `version`: a semantic version string; quote it (`\"1.0.0\"`). |\n| `description` | Optional text. |\n| `inputs` | Optional `{ schema }`: the run's static inputs must match it. |\n| `outputSchema` | Optional: the output of a `succeeded` terminal must match it. |\n| `initial` | The first state; not a terminal. |\n| `initialInput` | Optional JSON template (bindings: `inputs`, `kv`); the first state's input. Default null. |\n| `limits` | Optional `{ maxNodeExecutions?, timeoutSeconds?, onExceeded? }`. `maxNodeExecutions`: 1 to 10,000 per run, default 1,000. `timeoutSeconds`: the run's deadline, 1 to 2,592,000 seconds from when it begins; pause does not stop it. `onExceeded`: `expired` (the only policy). |\n| `catchAll` | Optional state that takes a runtime error no route takes; not `initial`; at most once per run. |\n| `schemas` | Named JSON Schema 2020-12 documents; refer to one as `{ $ref: \"#\u002Fschemas\u002FName\" }`. |\n| `contracts` | `Name@N: { schema, description?, response?: { schema, description? } }`. A request's contract needs `response`. |\n| `states` | 1 to 100 states, body states included. |\n\nAny schema field takes a `$ref` to `schemas` or an inline schema.\n\n## Bindings\n\nA binding is one whole string `\"${root.path}\"` or `\"${root.path ?? \u003CJSON literal>}\"`.\nRoots: `inputs`, `input`, `kv`, `runtime` (`runtime.error`), `branch`\n(`branch.item`, `branch.index`; `branch.name` in named branches; branch\nstates only). Path segments: `.name`,\n`[index]`, `[\"any key\"]`. A template is any JSON value whose strings may be\nbindings.\n\nA template is a YAML value: an object is a map (`{ fact: \"...\" }`), not\nJSON text in quotes (`'{\"fact\": \"...\"}'` is one string). A template with no\nbinding is checked against its contract or `inputSchema` when the draft is\nsaved (`WORKFLOW_STATE_INPUT_INVALID`); JSON written as text where an object\nor array is expected gets the warning `WORKFLOW_INPUT_JSON_TEXT`. A state\nwith no `input:` whose closed `inputSchema` refuses a property that the\n`outputSchema` of the state before it declares (a wait passes its input on),\nor needs one it lacks, gets the warning `WORKFLOW_INPUT_MISMATCH`.\n\n## States\n\nEvery state except a terminal has `transition: { next: { \u003Croute>: \u003Cstate> } }`.\nA route is an outcome of the state, a runtime error code or `failure` (any\nruntime error that has no route of its own). All states take an optional\n`description`.\n\n### script\n\n| Field | Rule |\n|---|---|\n| `source` | Required. Node ES module text. |\n| `runtime` | Optional; `node` only. |\n| `timeoutSeconds` | 1 to 3,600; default 300. |\n| `input` | Optional template; default: the value that arrived. |\n| `inputSchema`, `outputSchema` | Optional. |\n| `outcomes` | Optional list; default: its routes that are not runtime routes. |\n\nOutput: the `setResult` output. Errors: `invalid_input`, `invalid_output`\n(undeclared outcome, non-JSON or mismatched output), `execution_failed`\n(non-zero exit without a result), `timed_out`, `cancelled`.\n\n### wait\n\n| Field | Rule |\n|---|---|\n| `seconds` | 1 to 86,400, or one binding (checked at entry: `invalid_input`). Required unless `at`. |\n| `at` | Instead of `seconds`: `{ minute, hour?, day_of_week? (0-6, Sunday 0) or day_of_month? (1-31), tz }`; a day needs `hour`, and `minute` alone is every hour. Waits until the next such time in the IANA zone `tz` (daylight saving included), decided at entry. |\n| `input` | Optional template. |\n\nOutcome: `elapsed` (route required). Output: its input.\n\n### messaging, mode receive\n\n| Field | Rule |\n|---|---|\n| `contract` | Required. A root contract. |\n| `from` | Optional: `{ parent: true }`, a session id, or `{ sessionId }` (may be a binding). Without it, any permitted sender. |\n| `correlate` | Optional template object; a message's `correlation` must equal it. |\n| `wait` | Optional `{ seconds }` (as a wait state's seconds): the receive races a durable timer. |\n| `input` | Optional template. |\n\nOutcomes: `received` (output: the message payload) and, with `wait`,\n`elapsed` (output: its input). Routes for both are required. A message given\nbefore the state starts is taken too, in the order the run accepted it.\n\n### messaging, mode send\n\n| Field | Rule |\n|---|---|\n| `contract` | Required. |\n| `recipient` | Required: `{ parent: true }`, a session id, or `{ sessionId }` (may be a binding). |\n| `correlate` | Optional template object. |\n| `input` | Optional template; the message. At most 16 KB, valid for the contract. |\n\nOutcome: `sent` (output: the delivery receipt): the recipient accepted the\nmessage. The next state gets this receipt, not the message. That does not\nmean it was read. An agent session gets it in its\nqueue (at least once). A workflow gets it as typed input (once). Errors:\n`unavailable`, `forbidden`, `unsupported_contract`, `delivery_failed`,\n`invalid_input`.\n\n### messaging, mode request\n\n| Field | Rule |\n|---|---|\n| `contract` | Required. Its contract declares `response`. |\n| `recipient` | Required, as for send. |\n| `timeoutSeconds` | Required. 1 to 86,400, or one binding. |\n| `correlate` | Optional template object. |\n| `input` | Optional template; the request payload. |\n\nOutcomes: `responded` (output: the response, valid for the response schema)\nand `declined` (output: `{ reason }`). The next state gets only that output:\nnothing from before the request. Routes for both are required. The\ndeadline is `timed_out`. The recipient answers with\n`respond_session_request`; session is not an answer.\n\n### agent\n\n| Field | Rule |\n|---|---|\n| `invoke` | Required. `{ prompt: \u003Ctext, at most 32 KB> }` or `{ agent: { name, package?, version?, interface? } }`; `version` and `interface` need `package`. |\n| `timeoutSeconds` | Required. 1 to 86,400, or one binding. |\n| `input` | Optional template; checked against `inputSchema`. |\n| `inputSchema`, `outputSchema` | Optional. |\n| `outcomes` | Optional list of business outcomes; default: its routes that are not runtime routes. |\n| `session` | Optional. `continue` (default) or `new`. In a parallel body (For each or branch) only `new`, its default there. (`session` is no longer a field.) |\n| `tools` | Optional. `none` or tool names: all the agent and its sub-agents get (plus `complete_workflow_node`). |\n\nOutput: the agent's validated output. The agent may report `needs_input` or\n`execution_failed`. The workflow decides `timed_out`, `invalid_output` (three\ninvalid completions), `unavailable`, `forbidden` and `cancelled`.\n\n`session: continue` (the default) keeps one agent session across the state's\nvisits in a Workflow: each later visit arrives as \"cycle N\" with the input,\nthe variables now, which changed, and the previous result. The step\nremembers earlier cycles (no repeats, incremental work); its context grows,\nso it costs more. Use `session: new` (a fresh agent session each visit: clean,\ncheaper, no drift) only when the person asks for it, or the visits are\nindependent and the person wants that. In a For each or a parallel branch\neach visit has its own agent session: `new` is the default there and\n`continue` is refused. Another Workflow (Run again) starts new agent\nsessions. Each completion gives `node_execution_id` from its prompt.\nSession lifecycle: the time limit is per visit; the agent session ends with\nthe Workflow (a stop ends an idle one at once). An agent session that ended\non its own is replaced, and the prompt says so.\n\n`interface: \u003Cname>@\u003Cversion>` (for example `review_change@1`) names the\nworkflow interface its package agent exports. `ps_list_agents` lists it as\nthe entry's `workflowInterface` (`ref`, `outcomes`, `inputSchema`,\n`outputSchema`, `digest`); set `package` to the entry's `namespace`. The\nstate must declare `outputSchema`. At publication and at start,\nthe state's outcomes and `outputSchema` must equal the interface's,\nand `inputSchema` too when the interface has one. Copy each interface\nschema into `schemas` unchanged (no `$ref` inside the copy) and refer to it\nfrom the state; a stricter or looser schema does not match. A\nmismatch refuses with `CONTRACT_MISMATCH` and its `problems`; a match pins\nthe interface's digest with the package revision.\n\n### parallel\n\n| Field | Rule |\n|---|---|\n| `forEach` | One binding that gives a list, or a list; at most 64 items. |\n| `branches` | Instead of `forEach` and `body`: `{ \u003Cname>: { initial, states, input? } }`, 1 to 64, each its own graph from its `initial`, with its `input` (default: the value that arrived). `concurrency` defaults to their number; above 16, set it. |\n| `concurrency` | 1 to 16; default 1. |\n| `join` | Optional: `all` (default; `all_required` is the same) or `{ first: k }`, k from 1 to 64 and at most the items of a literal list. |\n| `failFast` | Optional boolean; default false. Cancels the other branches once the join can no longer succeed. |\n| `timeoutSeconds` | Optional: the state's own time limit, 1 to 86400 or one binding. Needs a `timed_out` route. |\n| `inputSchema` | Optional: checks the whole list. |\n| `outputSchema` | Optional: checks the joined output. |\n| `body` | With `forEach`: `{ initial, states }`. No parallel state inside. |\n\nOutcomes: `joined` and `blocked`. Routes for both are required.\n\n- `joined` with `all`: the branch outputs in list order. With `{ first: k }`:\n  `[{ index, output }]` of the first k successes, in the order the store\n  saved them (ties: lower index first); the other branches are cancelled or\n  not started. Named branches: `{ \u003Cname>: output }`, the winners only; a\n  report's branches add `name`. Without `timeoutSeconds`, a named branch\n  with an agent state, or a receive with no wait, warns\n  `WORKFLOW_BRANCH_UNBOUNDED`.\n- `blocked`: k successes can no longer happen (with `{ first: k }` and fewer\n  than k items, at once). Output: `{ joined: false, timedOut, total, needed,\n  succeeded, failFast, branches: [{ index, ended, nodeExecutionId, outcome,\n  resultKind, output or error, cancelledBy?, late? }] }`. `ended` is\n  `terminal`, `error`, `cancelled`, `not_started` or `finished` (its last\n  step's result was saved, and kept, but its next step never started). `cancelledBy` is `failFast`, `join` or `timeout`. `late` marks a\n  success saved after the time limit. A report too large to store drops\n  every `output` and `error` and adds `truncated: true`.\n- With `timeoutSeconds`: a success counts only when it was saved at or\n  before the time limit. A branch step whose own deadline\n  equals the time limit ends just after it, so it does not count. When the time limit passes with k such successes,\n  the state still joins. With fewer, it ends with the runtime error\n  `timed_out`, whose `details` is the same report with `timedOut: true`;\n  if k could already no longer happen then, it is `blocked` with\n  `timedOut: true` instead.\n- `outputSchema` checks only the joined output, never a report.\n- Worker slots: each running script holds one of the worker's slots\n  (`PILOTSWARM_WORKER_CONCURRENCY`; the duroxide default is 2) for its whole\n  run, and the controller needs one too: keep `concurrency` below the slot\n  count, with room to spare. Stopping a script is best-effort.\n\n### terminal\n\n| Field | Rule |\n|---|---|\n| `outcome` | Required: `succeeded`, `blocked`, `failed` or `cancelled`. |\n| `output` | Optional template; default null. |\n| `summary` | Optional text. |\n| `outputSchema` | Optional, in addition to the root `outputSchema` for `succeeded`. |\n\n## Runtime error codes\n\n`invalid_input`, `invalid_output`, `timed_out`, `cancelled`, `forbidden`,\n`unavailable`, `unsupported_contract`, `delivery_failed`, `budget_exceeded`,\n`needs_input`, `execution_failed`. On an error route, `input` and\n`runtime.error` are the envelope `{ code, message, details? }`.\n\n## Semantics\n\nWhat the engine guarantees. Do not rely on more.\n\n- **Durable.** The store keeps the run, each visit (input, outcome,\n  output), the variables, accepted messages and requests; the controller's\n  history keeps its place. After a worker restart the run goes on where\n  it was.\n- **Order.** The run starts at `initial`. Outside a parallel state one state\n  runs at a time: its outcome, or its runtime error, takes one route, and the\n  next state gets its output (or the error envelope). Every visit counts\n  toward `maxNodeExecutions`; at the limit the run fails (`limit_exceeded`).\n- **One result per visit.** The first result stored stands. Nothing is\n  undone: side effects that happened stay.\n\n| Effect | On a retry or a worker restart |\n|---|---|\n| script | Runs again from its start (`context.attempt` above 1), tool calls too: make it safe to repeat. |\n| agent | Keeps the same agent session; its prompt is not given twice. The first accepted `complete_workflow_node` decides; later ones are refused. |\n| send | An agent session may get it twice (a repeat is dropped while its id is recent). A workflow takes it once. |\n| request | Reaches the responder at least once, under one id. The first answer decides; none after the deadline. |\n| receive | Takes each message once, in the order the run accepted it. |\n| wait | Elapses once, by the store's clock. |\n\n- **Timers.** Every deadline (a wait, a receive's `wait`, a request, an agent\n  or parallel `timeoutSeconds`, the run's `limits.timeoutSeconds`) is a\n  durable timer, fixed when its state or run starts. Restarts and pauses do\n  not stop it.\n- **Joins.** `forEach` is read once, at entry. At most `concurrency`\n  branches run at once. Only successes stored in time count. A cancelled\n  branch's running script may still finish (stopping is best-effort), but\n  its result does not count.\n- **Pause.** No state starts and no result is routed until `resume`. Started\n  work finishes and keeps its result; a state whose deadline passes\n  meanwhile is routed after the resume.\n- **Stop.** Nothing new starts; started work settles. The run ends `stopped`\n  when nothing is pending, or at the end of its grace, cancelling the rest.\n- **Cancel, delete.** The run ends when it is accepted: pending visits are\n  cancelled, then scripts are stopped and agent sessions ended. A delete tells\n  no one.\n- **Run deadline.** Pending work is cancelled and the run fails `timed_out`,\n  paused or not.\n- **Failure.** A runtime error takes its code's route, else `failure`, else\n  `catchAll` (once per run; a later route to it fails the run), else the\n  run fails. In a branch it ends the branch without success. A run whose\n  controller stopped shows `needsRecovery`; `recover` ends it failed.\n- **Versions.** A Workflow runs the revision it started with, with its\n  named agents' pinned versions; a later save never changes it. To run a\n  change, start the new revision as another Workflow. Only its variables\n  change (`set_vars`); a step reads them as of its start.\n\n## Not supported: do not write\n\n`transition.function`, `transition.delaySeconds` (use a wait state), join\nstates, a parallel state in a body, `inputSchema` or `outputSchema` on a\nreceive, `parentNotification`, the request field `request` and the route\n`timeout`, and text around a binding.\n\n## Tools\n\nFor every argument of the Workflow tools and the refusals you can fix, call\n`load_skill` with name `workflow-tools`. Use `read_workflow`, not\n`list_sessions`, for a Workflow's status.\n\n## Example: review fan-out with bounded rework\n\n```yaml\n# example: change-review\napiVersion: pilotswarm.dev\u002Fv1\nkind: Workflow\nmetadata: { name: change-review, version: \"1.0.0\" }\ndescription: Reviews pinned changes in parallel with bounded rework, then asks the parent to approve the reviewed set.\ninputs:\n  schema: { $ref: \"#\u002Fschemas\u002FReviewInputs\" }\noutputSchema: { $ref: \"#\u002Fschemas\u002FDecision\" }\ninitial: review_changes\ninitialInput: \"${inputs.changes}\"\nschemas:\n  Change:\n    type: object\n    additionalProperties: false\n    required: [workId, head, authorSessionId, round]\n    properties:\n      workId: { type: string, minLength: 1 }\n      head: { type: string, minLength: 40, maxLength: 40 }\n      authorSessionId: { type: string, minLength: 1 }\n      round: { type: integer, minimum: 1, maximum: 3 }\n  ReviewInputs:\n    type: object\n    additionalProperties: false\n    required: [changes]\n    properties:\n      changes: { type: array, minItems: 1, maxItems: 16, items: { $ref: \"#\u002Fschemas\u002FChange\" } }\n  ReviewResult:\n    type: object\n    additionalProperties: false\n    required: [workId, head, findings]\n    properties:\n      workId: { type: string }\n      head: { type: string }\n      findings: { type: array, items: { type: string } }\n  NewHead:\n    type: object\n    additionalProperties: false\n    required: [head]\n    properties:\n      head: { type: string, minLength: 40, maxLength: 40 }\n  ReviewedSet:\n    type: array\n    items: { $ref: \"#\u002Fschemas\u002FReviewResult\" }\n  Decision:\n    type: object\n    additionalProperties: false\n    required: [approved]\n    properties:\n      approved: { type: boolean }\n      note: { type: string }\ncontracts:\n  Rework@1:\n    schema: { $ref: \"#\u002Fschemas\u002FReviewResult\" }\n    response: { schema: { $ref: \"#\u002Fschemas\u002FNewHead\" } }\n  Approve@1:\n    schema: { $ref: \"#\u002Fschemas\u002FReviewedSet\" }\n    response: { schema: { $ref: \"#\u002Fschemas\u002FDecision\" } }\nstates:\n  review_changes:\n    type: parallel\n    forEach: \"${input}\"\n    concurrency: 4\n    outputSchema: { $ref: \"#\u002Fschemas\u002FReviewedSet\" }\n    body:\n      initial: review\n      states:\n        review:\n          type: agent\n          invoke: { agent: { name: code-reviewer, package: review-kit, version: \"1.2.0\" } }\n          inputSchema: { $ref: \"#\u002Fschemas\u002FChange\" }\n          outputSchema: { $ref: \"#\u002Fschemas\u002FReviewResult\" }\n          outcomes: [accepted, changes_required]\n          timeoutSeconds: 900\n          transition:\n            next: { accepted: accepted, changes_required: ask_rework, invalid_input: blocked, needs_input: blocked, failure: blocked }\n        ask_rework:\n          type: messaging\n          mode: request\n          contract: Rework@1\n          recipient: { sessionId: \"${branch.item.authorSessionId}\" }\n          timeoutSeconds: 3600\n          transition:\n            next: { responded: next_round, declined: blocked, failure: blocked }\n        next_round:\n          type: script\n          timeoutSeconds: 120\n          outcomes: [verified]\n          source: |\n            const item = context.branch.item;\n            const checked = await callTool(\"verify_review_round\", { workId: item.workId, head: context.input.head });\n            await setResult(\"verified\", { ...item, head: context.input.head, round: checked.round });\n          transition:\n            next: { verified: review, failure: blocked }\n        accepted:\n          type: terminal\n          outcome: succeeded\n          output: \"${input}\"\n        blocked:\n          type: terminal\n          outcome: blocked\n          output: \"${input}\"\n    transition:\n      next: { joined: approve, blocked: stuck, failure: failed }\n  approve:\n    type: messaging\n    mode: request\n    contract: Approve@1\n    recipient: { parent: true }\n    timeoutSeconds: 86400\n    transition:\n      next: { responded: decide, declined: rejected, timed_out: rejected, failure: failed }\n  decide:\n    type: script\n    outcomes: [approved, rejected]\n    source: |\n      await setResult(context.input.approved ? \"approved\" : \"rejected\", context.input);\n    transition:\n      next: { approved: done, rejected: rejected, failure: failed }\n  done:\n    type: terminal\n    outcome: succeeded\n    output: \"${input}\"\n  rejected:\n    type: terminal\n    outcome: blocked\n    output: \"${input}\"\n  stuck:\n    type: terminal\n    outcome: blocked\n    output: \"${input}\"\n  failed:\n    type: terminal\n    outcome: failed\n    output: \"${runtime.error}\"\n```\n\n`code-reviewer` in `review-kit` and `verify_review_round` stand for an agent\nand a tool of your own. The rework loop is bounded by the schema: a fourth\nround fails `review`'s input check (`invalid_input`) and ends the branch\nblocked; a blocked join ends at `stuck` with every branch's result.\n","# Workflow reference\n\nExact fields for `apiVersion: pilotswarm.dev\u002Fv1` definitions and the\nworkflow tools. `workflow-authoring` explains how to design with them.\n\nIds: a state id or an outcome is a letter, then at most 63 letters, digits,\n`_` or `-`. State ids are unique in the whole workflow, body and branch\nstates included. A contract name is `Name@N` (for example `BuildEvent@1`).\n\n## Root keys\n\n| Key | Rule |\n|---|---|\n| `apiVersion` | `pilotswarm.dev\u002Fv1` |\n| `kind` | `Workflow` |\n| `metadata` | `{ name, version, description? }`. `name`: at most 128 characters. `version`: a semantic version string; quote it (`\"1.0.0\"`). |\n| `description` | Optional text. |\n| `inputs` | Optional `{ schema }`: the run's static inputs must match it. |\n| `outputSchema` | Optional: the output of a `succeeded` terminal must match it. |\n| `initial` | The first state; not a terminal. |\n| `initialInput` | Optional JSON template (bindings: `inputs`, `kv`); the first state's input. Default null. |\n| `limits` | Optional `{ maxNodeExecutions?, timeoutSeconds?, onExceeded? }`. `maxNodeExecutions`: 1 to 10,000 per run, default 1,000. `timeoutSeconds`: the run's deadline, 1 to 2,592,000 seconds from when it begins; pause does not stop it. `onExceeded`: `expired` (the only policy). |\n| `catchAll` | Optional state that takes a runtime error no route takes; not `initial`; at most once per run. |\n| `schemas` | Named JSON Schema 2020-12 documents; refer to one as `{ $ref: \"#\u002Fschemas\u002FName\" }`. |\n| `contracts` | `Name@N: { schema, description?, response?: { schema, description? } }`. A request's contract needs `response`. |\n| `states` | 1 to 100 states, body states included. |\n\nAny schema field takes a `$ref` to `schemas` or an inline schema.\n\n## Bindings\n\nA binding is one whole string `\"${root.path}\"` or `\"${root.path ?? \u003CJSON literal>}\"`.\nRoots: `inputs`, `input`, `kv`, `runtime` (`runtime.error`), `branch`\n(`branch.item`, `branch.index`; `branch.name` in named branches; branch\nstates only). Path segments: `.name`,\n`[index]`, `[\"any key\"]`. A template is any JSON value whose strings may be\nbindings.\n\nA template is a YAML value: an object is a map (`{ fact: \"...\" }`), not\nJSON text in quotes (`'{\"fact\": \"...\"}'` is one string). A template with no\nbinding is checked against its contract or `inputSchema` when the draft is\nsaved (`WORKFLOW_STATE_INPUT_INVALID`); JSON written as text where an object\nor array is expected gets the warning `WORKFLOW_INPUT_JSON_TEXT`. A state\nwith no `input:` whose closed `inputSchema` refuses a property that the\n`outputSchema` of the state before it declares (a wait passes its input on),\nor needs one it lacks, gets the warning `WORKFLOW_INPUT_MISMATCH`.\n\n## States\n\nEvery state except a terminal has `transition: { next: { \u003Croute>: \u003Cstate> } }`.\nA route is an outcome of the state, a runtime error code or `failure` (any\nruntime error that has no route of its own). All states take an optional\n`description`.\n\n### script\n\n| Field | Rule |\n|---|---|\n| `source` | Required. Node ES module text. |\n| `runtime` | Optional; `node` only. |\n| `timeoutSeconds` | 1 to 3,600; default 300. |\n| `input` | Optional template; default: the value that arrived. |\n| `inputSchema`, `outputSchema` | Optional. |\n| `outcomes` | Optional list; default: its routes that are not runtime routes. |\n\nOutput: the `setResult` output. Errors: `invalid_input`, `invalid_output`\n(undeclared outcome, non-JSON or mismatched output), `execution_failed`\n(non-zero exit without a result), `timed_out`, `cancelled`.\n\n### wait\n\n| Field | Rule |\n|---|---|\n| `seconds` | 1 to 86,400, or one binding (checked at entry: `invalid_input`). Required unless `at`. |\n| `at` | Instead of `seconds`: `{ minute, hour?, day_of_week? (0-6, Sunday 0) or day_of_month? (1-31), tz }`; a day needs `hour`, and `minute` alone is every hour. Waits until the next such time in the IANA zone `tz` (daylight saving included), decided at entry. |\n| `input` | Optional template. |\n\nOutcome: `elapsed` (route required). Output: its input.\n\n### messaging, mode receive\n\n| Field | Rule |\n|---|---|\n| `contract` | Required. A root contract. |\n| `from` | Optional: `{ parent: true }`, a session id, or `{ sessionId }` (may be a binding). Without it, any permitted sender. |\n| `correlate` | Optional template object; a message's `correlation` must equal it. |\n| `wait` | Optional `{ seconds }` (as a wait state's seconds): the receive races a durable timer. |\n| `input` | Optional template. |\n\nOutcomes: `received` (output: the message payload) and, with `wait`,\n`elapsed` (output: its input). Routes for both are required. A message given\nbefore the state starts is taken too, in the order the run accepted it.\n\n### messaging, mode send\n\n| Field | Rule |\n|---|---|\n| `contract` | Required. |\n| `recipient` | Required: `{ parent: true }`, a session id, or `{ sessionId }` (may be a binding). |\n| `correlate` | Optional template object. |\n| `input` | Optional template; the message. At most 16 KB, valid for the contract. |\n\nOutcome: `sent` (output: the delivery receipt): the recipient accepted the\nmessage. The next state gets this receipt, not the message. That does not\nmean it was read. An agent session gets it in its\nqueue (at least once). A workflow gets it as typed input (once). Errors:\n`unavailable`, `forbidden`, `unsupported_contract`, `delivery_failed`,\n`invalid_input`.\n\n### messaging, mode request\n\n| Field | Rule |\n|---|---|\n| `contract` | Required. Its contract declares `response`. |\n| `recipient` | Required, as for send. |\n| `timeoutSeconds` | Required. 1 to 86,400, or one binding. |\n| `correlate` | Optional template object. |\n| `input` | Optional template; the request payload. |\n\nOutcomes: `responded` (output: the response, valid for the response schema)\nand `declined` (output: `{ reason }`). The next state gets only that output:\nnothing from before the request. Routes for both are required. The\ndeadline is `timed_out`. The recipient answers with\n`respond_session_request`; session is not an answer.\n\n### agent\n\n| Field | Rule |\n|---|---|\n| `invoke` | Required. `{ prompt: \u003Ctext, at most 32 KB> }` or `{ agent: { name, package?, version?, interface? } }`; `version` and `interface` need `package`. |\n| `timeoutSeconds` | Required. 1 to 86,400, or one binding. |\n| `input` | Optional template; checked against `inputSchema`. |\n| `inputSchema`, `outputSchema` | Optional. |\n| `outcomes` | Optional list of business outcomes; default: its routes that are not runtime routes. |\n| `session` | Optional. `continue` (default) or `new`. In a parallel body (For each or branch) only `new`, its default there. (`session` is no longer a field.) |\n| `tools` | Optional. `none` or tool names: all the agent and its sub-agents get (plus `complete_workflow_node`). |\n\nOutput: the agent's validated output. The agent may report `needs_input` or\n`execution_failed`. The workflow decides `timed_out`, `invalid_output` (three\ninvalid completions), `unavailable`, `forbidden` and `cancelled`.\n\n`session: continue` (the default) keeps one agent session across the state's\nvisits in a Workflow: each later visit arrives as \"cycle N\" with the input,\nthe variables now, which changed, and the previous result. The step\nremembers earlier cycles (no repeats, incremental work); its context grows,\nso it costs more. Use `session: new` (a fresh agent session each visit: clean,\ncheaper, no drift) only when the person asks for it, or the visits are\nindependent and the person wants that. In a For each or a parallel branch\neach visit has its own agent session: `new` is the default there and\n`continue` is refused. Another Workflow (Run again) starts new agent\nsessions. Each completion gives `node_execution_id` from its prompt.\nSession lifecycle: the time limit is per visit; the agent session ends with\nthe Workflow (a stop ends an idle one at once). An agent session that ended\non its own is replaced, and the prompt says so.\n\n`interface: \u003Cname>@\u003Cversion>` (for example `review_change@1`) names the\nworkflow interface its package agent exports. `ps_list_agents` lists it as\nthe entry's `workflowInterface` (`ref`, `outcomes`, `inputSchema`,\n`outputSchema`, `digest`); set `package` to the entry's `namespace`. The\nstate must declare `outputSchema`. At publication and at start,\nthe state's outcomes and `outputSchema` must equal the interface's,\nand `inputSchema` too when the interface has one. Copy each interface\nschema into `schemas` unchanged (no `$ref` inside the copy) and refer to it\nfrom the state; a stricter or looser schema does not match. A\nmismatch refuses with `CONTRACT_MISMATCH` and its `problems`; a match pins\nthe interface's digest with the package revision.\n\n### parallel\n\n| Field | Rule |\n|---|---|\n| `forEach` | One binding that gives a list, or a list; at most 64 items. |\n| `branches` | Instead of `forEach` and `body`: `{ \u003Cname>: { initial, states, input? } }`, 1 to 64, each its own graph from its `initial`, with its `input` (default: the value that arrived). `concurrency` defaults to their number; above 16, set it. |\n| `concurrency` | 1 to 16; default 1. |\n| `join` | Optional: `all` (default; `all_required` is the same) or `{ first: k }`, k from 1 to 64 and at most the items of a literal list. |\n| `failFast` | Optional boolean; default false. Cancels the other branches once the join can no longer succeed. |\n| `timeoutSeconds` | Optional: the state's own time limit, 1 to 86400 or one binding. Needs a `timed_out` route. |\n| `inputSchema` | Optional: checks the whole list. |\n| `outputSchema` | Optional: checks the joined output. |\n| `body` | With `forEach`: `{ initial, states }`. No parallel state inside. |\n\nOutcomes: `joined` and `blocked`. Routes for both are required.\n\n- `joined` with `all`: the branch outputs in list order. With `{ first: k }`:\n  `[{ index, output }]` of the first k successes, in the order the store\n  saved them (ties: lower index first); the other branches are cancelled or\n  not started. Named branches: `{ \u003Cname>: output }`, the winners only; a\n  report's branches add `name`. Without `timeoutSeconds`, a named branch\n  with an agent state, or a receive with no wait, warns\n  `WORKFLOW_BRANCH_UNBOUNDED`.\n- `blocked`: k successes can no longer happen (with `{ first: k }` and fewer\n  than k items, at once). Output: `{ joined: false, timedOut, total, needed,\n  succeeded, failFast, branches: [{ index, ended, nodeExecutionId, outcome,\n  resultKind, output or error, cancelledBy?, late? }] }`. `ended` is\n  `terminal`, `error`, `cancelled`, `not_started` or `finished` (its last\n  step's result was saved, and kept, but its next step never started). `cancelledBy` is `failFast`, `join` or `timeout`. `late` marks a\n  success saved after the time limit. A report too large to store drops\n  every `output` and `error` and adds `truncated: true`.\n- With `timeoutSeconds`: a success counts only when it was saved at or\n  before the time limit. A branch step whose own deadline\n  equals the time limit ends just after it, so it does not count. When the time limit passes with k such successes,\n  the state still joins. With fewer, it ends with the runtime error\n  `timed_out`, whose `details` is the same report with `timedOut: true`;\n  if k could already no longer happen then, it is `blocked` with\n  `timedOut: true` instead.\n- `outputSchema` checks only the joined output, never a report.\n- Worker slots: each running script holds one of the worker's slots\n  (`PILOTSWARM_WORKER_CONCURRENCY`; the duroxide default is 2) for its whole\n  run, and the controller needs one too: keep `concurrency` below the slot\n  count, with room to spare. Stopping a script is best-effort.\n\n### terminal\n\n| Field | Rule |\n|---|---|\n| `outcome` | Required: `succeeded`, `blocked`, `failed` or `cancelled`. |\n| `output` | Optional template; default null. |\n| `summary` | Optional text. |\n| `outputSchema` | Optional, in addition to the root `outputSchema` for `succeeded`. |\n\n## Runtime error codes\n\n`invalid_input`, `invalid_output`, `timed_out`, `cancelled`, `forbidden`,\n`unavailable`, `unsupported_contract`, `delivery_failed`, `budget_exceeded`,\n`needs_input`, `execution_failed`. On an error route, `input` and\n`runtime.error` are the envelope `{ code, message, details? }`.\n\n## Semantics\n\nWhat the engine guarantees. Do not rely on more.\n\n- **Durable.** The store keeps the run, each visit (input, outcome,\n  output), the variables, accepted messages and requests; the controller's\n  history keeps its place. After a worker restart the run goes on where\n  it was.\n- **Order.** The run starts at `initial`. Outside a parallel state one state\n  runs at a time: its outcome, or its runtime error, takes one route, and the\n  next state gets its output (or the error envelope). Every visit counts\n  toward `maxNodeExecutions`; at the limit the run fails (`limit_exceeded`).\n- **One result per visit.** The first result stored stands. Nothing is\n  undone: side effects that happened stay.\n\n| Effect | On a retry or a worker restart |\n|---|---|\n| script | Runs again from its start (`context.attempt` above 1), tool calls too: make it safe to repeat. |\n| agent | Keeps the same agent session; its prompt is not given twice. The first accepted `complete_workflow_node` decides; later ones are refused. |\n| send | An agent session may get it twice (a repeat is dropped while its id is recent). A workflow takes it once. |\n| request | Reaches the responder at least once, under one id. The first answer decides; none after the deadline. |\n| receive | Takes each message once, in the order the run accepted it. |\n| wait | Elapses once, by the store's clock. |\n\n- **Timers.** Every deadline (a wait, a receive's `wait`, a request, an agent\n  or parallel `timeoutSeconds`, the run's `limits.timeoutSeconds`) is a\n  durable timer, fixed when its state or run starts. Restarts and pauses do\n  not stop it.\n- **Joins.** `forEach` is read once, at entry. At most `concurrency`\n  branches run at once. Only successes stored in time count. A cancelled\n  branch's running script may still finish (stopping is best-effort), but\n  its result does not count.\n- **Pause.** No state starts and no result is routed until `resume`. Started\n  work finishes and keeps its result; a state whose deadline passes\n  meanwhile is routed after the resume.\n- **Stop.** Nothing new starts; started work settles. The run ends `stopped`\n  when nothing is pending, or at the end of its grace, cancelling the rest.\n- **Cancel, delete.** The run ends when it is accepted: pending visits are\n  cancelled, then scripts are stopped and agent sessions ended. A delete tells\n  no one.\n- **Run deadline.** Pending work is cancelled and the run fails `timed_out`,\n  paused or not.\n- **Failure.** A runtime error takes its code's route, else `failure`, else\n  `catchAll` (once per run; a later route to it fails the run), else the\n  run fails. In a branch it ends the branch without success. A run whose\n  controller stopped shows `needsRecovery`; `recover` ends it failed.\n- **Versions.** A Workflow runs the revision it started with, with its\n  named agents' pinned versions; a later save never changes it. To run a\n  change, start the new revision as another Workflow. Only its variables\n  change (`set_vars`); a step reads them as of its start.\n\n## Not supported: do not write\n\n`transition.function`, `transition.delaySeconds` (use a wait state), join\nstates, a parallel state in a body, `inputSchema` or `outputSchema` on a\nreceive, `parentNotification`, the request field `request` and the route\n`timeout`, and text around a binding.\n\n## Tools\n\nFor every argument of the Workflow tools and the refusals you can fix, call\n`load_skill` with name `workflow-tools`. Use `read_workflow`, not\n`list_sessions`, for a Workflow's status.\n\n## Example: review fan-out with bounded rework\n\n```yaml\n# example: change-review\napiVersion: pilotswarm.dev\u002Fv1\nkind: Workflow\nmetadata: { name: change-review, version: \"1.0.0\" }\ndescription: Reviews pinned changes in parallel with bounded rework, then asks the parent to approve the reviewed set.\ninputs:\n  schema: { $ref: \"#\u002Fschemas\u002FReviewInputs\" }\noutputSchema: { $ref: \"#\u002Fschemas\u002FDecision\" }\ninitial: review_changes\ninitialInput: \"${inputs.changes}\"\nschemas:\n  Change:\n    type: object\n    additionalProperties: false\n    required: [workId, head, authorSessionId, round]\n    properties:\n      workId: { type: string, minLength: 1 }\n      head: { type: string, minLength: 40, maxLength: 40 }\n      authorSessionId: { type: string, minLength: 1 }\n      round: { type: integer, minimum: 1, maximum: 3 }\n  ReviewInputs:\n    type: object\n    additionalProperties: false\n    required: [changes]\n    properties:\n      changes: { type: array, minItems: 1, maxItems: 16, items: { $ref: \"#\u002Fschemas\u002FChange\" } }\n  ReviewResult:\n    type: object\n    additionalProperties: false\n    required: [workId, head, findings]\n    properties:\n      workId: { type: string }\n      head: { type: string }\n      findings: { type: array, items: { type: string } }\n  NewHead:\n    type: object\n    additionalProperties: false\n    required: [head]\n    properties:\n      head: { type: string, minLength: 40, maxLength: 40 }\n  ReviewedSet:\n    type: array\n    items: { $ref: \"#\u002Fschemas\u002FReviewResult\" }\n  Decision:\n    type: object\n    additionalProperties: false\n    required: [approved]\n    properties:\n      approved: { type: boolean }\n      note: { type: string }\ncontracts:\n  Rework@1:\n    schema: { $ref: \"#\u002Fschemas\u002FReviewResult\" }\n    response: { schema: { $ref: \"#\u002Fschemas\u002FNewHead\" } }\n  Approve@1:\n    schema: { $ref: \"#\u002Fschemas\u002FReviewedSet\" }\n    response: { schema: { $ref: \"#\u002Fschemas\u002FDecision\" } }\nstates:\n  review_changes:\n    type: parallel\n    forEach: \"${input}\"\n    concurrency: 4\n    outputSchema: { $ref: \"#\u002Fschemas\u002FReviewedSet\" }\n    body:\n      initial: review\n      states:\n        review:\n          type: agent\n          invoke: { agent: { name: code-reviewer, package: review-kit, version: \"1.2.0\" } }\n          inputSchema: { $ref: \"#\u002Fschemas\u002FChange\" }\n          outputSchema: { $ref: \"#\u002Fschemas\u002FReviewResult\" }\n          outcomes: [accepted, changes_required]\n          timeoutSeconds: 900\n          transition:\n            next: { accepted: accepted, changes_required: ask_rework, invalid_input: blocked, needs_input: blocked, failure: blocked }\n        ask_rework:\n          type: messaging\n          mode: request\n          contract: Rework@1\n          recipient: { sessionId: \"${branch.item.authorSessionId}\" }\n          timeoutSeconds: 3600\n          transition:\n            next: { responded: next_round, declined: blocked, failure: blocked }\n        next_round:\n          type: script\n          timeoutSeconds: 120\n          outcomes: [verified]\n          source: |\n            const item = context.branch.item;\n            const checked = await callTool(\"verify_review_round\", { workId: item.workId, head: context.input.head });\n            await setResult(\"verified\", { ...item, head: context.input.head, round: checked.round });\n          transition:\n            next: { verified: review, failure: blocked }\n        accepted:\n          type: terminal\n          outcome: succeeded\n          output: \"${input}\"\n        blocked:\n          type: terminal\n          outcome: blocked\n          output: \"${input}\"\n    transition:\n      next: { joined: approve, blocked: stuck, failure: failed }\n  approve:\n    type: messaging\n    mode: request\n    contract: Approve@1\n    recipient: { parent: true }\n    timeoutSeconds: 86400\n    transition:\n      next: { responded: decide, declined: rejected, timed_out: rejected, failure: failed }\n  decide:\n    type: script\n    outcomes: [approved, rejected]\n    source: |\n      await setResult(context.input.approved ? \"approved\" : \"rejected\", context.input);\n    transition:\n      next: { approved: done, rejected: rejected, failure: failed }\n  done:\n    type: terminal\n    outcome: succeeded\n    output: \"${input}\"\n  rejected:\n    type: terminal\n    outcome: blocked\n    output: \"${input}\"\n  stuck:\n    type: terminal\n    outcome: blocked\n    output: \"${input}\"\n  failed:\n    type: terminal\n    outcome: failed\n    output: \"${runtime.error}\"\n```\n\n`code-reviewer` in `review-kit` and `verify_review_round` stand for an agent\nand a tool of your own. The rework loop is bounded by the schema: a fourth\nround fails `review`'s input check (`invalid_input`) and ends the branch\nblocked; a blocked join ends at `stuck` with every branch's result.\n",{"name":5,"description":7},"main","7ee6eff4333dba9512363bc67d1fe177d9568636","packages\u002Fsdk\u002Fplugins\u002Fworkflow\u002Fskills\u002Fworkflow-reference",10,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FPilotSwarm","https:\u002F\u002Fgithub.com\u002Fmicrosoft","https:\u002F\u002Fopensource.microsoft.com","Open source projects and samples from Microsoft","2026-11-15 00:36:04",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add microsoft\u002FPilotSwarm --skill workflow-reference",{"tool":39,"command":40},"bunx","bunx skills add microsoft\u002FPilotSwarm --skill workflow-reference",{"tool":42,"command":43},"pnpm","pnpm dlx skills add microsoft\u002FPilotSwarm --skill workflow-reference","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FPilotSwarm\u002Fblob\u002Fmain\u002Fpackages\u002Fsdk\u002Fplugins\u002Fworkflow\u002Fskills\u002Fworkflow-reference\u002FSKILL.md",true,[],[48,55,62,69,76,83],{"skillNo":49,"name":50,"nameZh":6,"description":51,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":52,"categories":53,"keywords":54,"publishedAt":19},"01M760N7TCZBTY635MYZQ8YAB0","workflow-testing","Load before you test a Workflow with mocked effects — the test, test_respond and test_advance operations, fixtures, interception ids, valid answers, the virtual clock, test evidence and what a test run does not prove.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FPilotSwarm\u002Fblob\u002Fmain\u002Fpackages\u002Fsdk\u002Fplugins\u002Fworkflow\u002Ftest-mode-skills\u002Fworkflow-testing\u002FSKILL.md",[],[],{"skillNo":56,"name":57,"nameZh":6,"description":58,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":19},"01M760N4RKD6BMYPGHD490SNN2","workflow-tools","Every argument of the Workflow tools, testing a revision, package definitions and the refusals you can fix. Load when a Workflow tool's short description is not enough.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FPilotSwarm\u002Fblob\u002Fmain\u002Fpackages\u002Fsdk\u002Fplugins\u002Fworkflow\u002Fskills\u002Fworkflow-tools\u002FSKILL.md",[],[],{"skillNo":63,"name":64,"nameZh":6,"description":65,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":19},"01M760MZ5CFSAX4FYMEXZJBGM2","workflow-patterns","Load when one Workflow must run different nodes in its branches at the same time (a different prompt or model for each branch), or loops back to a state (a polling loop). Named branches, the router pattern, examples that compile.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FPilotSwarm\u002Fblob\u002Fmain\u002Fpackages\u002Fsdk\u002Fplugins\u002Fworkflow\u002Fskills\u002Fworkflow-patterns\u002FSKILL.md",[],[],{"skillNo":70,"name":71,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":19},"01M760MWKBDJFEECRFAF3BAVQ2","workflow-authoring","Load before you create, edit, publish or run a Workflow (pilotswarm.dev\u002Fv1) — choosing states, bounded graphs, bindings, typed contracts, agent and parallel states, the lifecycle and failure diagnosis.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FPilotSwarm\u002Fblob\u002Fmain\u002Fpackages\u002Fsdk\u002Fplugins\u002Fworkflow\u002Fskills\u002Fworkflow-authoring\u002FSKILL.md",[],[],{"skillNo":77,"name":78,"nameZh":6,"description":79,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":19},"01M760MN4Y9199ARAY85DWD08X","tla-formal-validation","Required for any change to the core orchestrations (the durable session orchestration, session hydration and dehydration, and the Workflow controller and store). Teaches how to model the change in TLA+, update the existing models, run TLC with the repo runner, read the results, turn a counterexample into a fix and a regression test, and record the binding to the code.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FPilotSwarm\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Ftla-formal-validation\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":105},"01M69A9CY8CXHED3AFC8DAT9G0","pilotswarm-three-tier","Reference for adding a Tier 3 worker AKS cluster to a PilotSwarm deployment. Covers provisioning, identity, Dockerfile changes, entrypoint, manifests, and verification.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FPilotSwarm\u002Fblob\u002Fmain\u002Ftemplates\u002Fbuilder-agents\u002Fskills\u002Fpilotswarm-three-tier\u002FSKILL.md",[],[90,91,92,93,94,95,96,97,98,99,100,101,102,103,104],"pilotswarm three tier","three","identity","cluster","for","deployment","dockerfile","worker","manifests","verification","changes","aks","entrypoint","provisioning","reference","2026-11-10 10:12:30"]