Classify a single operator WorkIntent's signal vector (coupling, context pressure, skill boundary, review independence, budget, operator burden) into exactly one of the seven topology archetypes (node, scout, chain, DAG-workgroup, tournament, ambient-watcher, human-gate), and audit that legacy launch verbs (spawn/dispatch/sortie/conjure/nightshift) stay compatibility metadata instead of writing independent Agent Node/session/transcript state. Use when routing a new WorkIntent to a topology before any Agent Node is materialized, auditing whether a launch path secretly opens parallel governance state, or reviewing a control-plane PR that adds a new legacy-verb compatibility shim. NOT for the inter-agent protocol mechanics once a workgroup or tournament is actually summoned (use swarm-invocation-designer), decomposing the work into sub-tasks or building/scheduling the DAG graph after a DAG-workgroup archetype is chosen (use task-decomposer or next-move), picking which specific skill or subagent fills a chosen no
Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Classify a single operator WorkIntent's signal vector (coupling, context pressure, skill boundary, review independence, budget, operator burden) into exactly one of the seven topology archetypes (node, scout, chain, DAG-workgroup, tournament, ambient-watcher, human-gate), and audit that legacy launch verbs (spawn/dispatch/sortie/conjure/nightshift) stay compatibility metadata instead of writing independent Agent Node/session/transcript state. Use when routing a new WorkIntent to a topology before any Agent Node is materialized, auditing whether a launch path secretly opens parallel governance state, or reviewing a control-plane PR that adds a new legacy-verb compatibility shim. NOT for the inter-agent protocol mechanics once a workgroup or tournament is actually summoned (use swarm-invocation-designer), decomposing the work into sub-tasks or building/scheduling the DAG graph after a DAG-workgroup archetype is chosen (use task-decomposer or next-move), picking which specific skill or subagent fills a chosen node (use skillful-subagent-creator), or designing the approval mechanics of a chosen human-gate archetype (use human-gate-designer).
license
Apache-2.0
allowed-tools
Read,Write,Edit,Bash,Grep,Glob
metadata
{"category":"Agent & Orchestration","tags":["work-intake","topology-archetype","single-operator-action","legacy-verb-compat","agent-node-governance"],"provenance":{"kind":"first-party","owners":["port-daddy"]},"pairs-with":[{"skill":"swarm-invocation-designer","reason":"Once a chain/dag-workgroup/tournament archetype is chosen, this skill hands off the actual inter-agent invocation and coordination mechanics to it."},{"skill":"task-decomposer","reason":"A dag-workgroup or chain archetype still needs its sub-tasks decomposed; this skill only proves the archetype call itself is singular and legacy-safe."},{"skill":"next-move","reason":"next-move's synthesizer picks a planning/runtime topology per subtask; this skill is the upstream gate proving the operator-level WorkIntent resolved to one archetype before that DAG is even built."}],"io-contract":{"kind":"deliverable","consumes":["[Truncated]","[Truncated]"],"produces":["[Truncated]","[Truncated]"]}}
Work Intake Node Shaping
Route one operator WorkIntent to exactly one topology archetype, and keep legacy launch verbs as compatibility metadata instead of a second, ungoverned path to an Agent Node.
Use This For
Classifying a new WorkIntent's signal vector into exactly one of the seven topology archetypes before any Agent Node is materialized.
Auditing a spawn/dispatch/sortie/conjure/nightshift compatibility shim to confirm it routes into the shared WorkIntent -> WorkPlan pipeline instead of writing its own session/transcript state.
Reviewing a control-plane PR that adds a new launch entrypoint, to catch a second, independent archetype decision hiding behind a different verb.
Deciding whether ambiguous work (a "quick fix" vs. a "long research chain" vs. a "recurring watcher") should be a node, scout, or ambient-watcher before committing budget and skills to it.
Grading a proposed WorkIntent-to-archetype mapping against the "exactly one archetype" invariant before it reaches the daemon.
Do Not Use This For
Inter-agent protocol/IPC mechanics once a workgroup or tournament is actually summoned (swarm-invocation-designer).
Decomposing the work into sub-tasks or building/scheduling the DAG graph after a DAG-workgroup archetype is chosen (task-decomposer, next-move).
Picking which specific skill or subagent fills a chosen node, or designing a chosen human-gate's approval mechanics (skillful-subagent-creator, human-gate-designer).
Decision Model
flowchart TD
A[Operator WorkIntent + six-signal vector] --> B{Score coupling, context pressure, skill boundary, review independence, budget, operator burden}
B --> C{Exactly one archetype fits?}
C -->|Zero or many tie| D[Reject: rescore the signal vector, do not launch multiple]
C -->|Yes| E[Select one: node / scout / chain / dag-workgroup / tournament / ambient-watcher / human-gate]
E --> F{Legacy verb reachable? spawn/dispatch/sortie/conjure/nightshift}
F -->|Yes| G{Does it write its own Agent Node/session/transcript?}
G -->|Yes| H[FAIL: legacy verb bypasses the shared WorkPlan pipeline]
G -->|No| I[OK: legacy verb is compatibility metadata only]
F -->|No| I
I --> J[Materialize exactly one governed Agent Node]
Capture the WorkIntent's six-signal vector — coupling, context pressure, skill boundary, review independence, budget, operator burden. See references/seven-archetypes.md for what each signal measures.
Map the signal vector to exactly one of the seven canonical archetypes using the disambiguation heuristics in references/seven-archetypes.md. If two archetypes seem to fit, that is a scoring bug, not a tie to break by picking both.
Record selectedArchetypes as a single-element array. A WorkIntent that produces zero or multiple archetypes must be rejected before any Agent Node is materialized.
Enumerate every legacy launch verb (spawn, dispatch, sortie, conjure, nightshift) reachable from this WorkIntent's entrypoint and trace whether each one independently opens its own Agent Node, session id, or transcript instead of routing through the shared WorkIntent -> WorkPlan -> Agent Node pipeline. See references/legacy-verb-compatibility.md.
Flag any legacy route that writes independent state. It means an old code path is quietly bypassing the daemon's governance predicates (session, worktree, transcript chain, cost cap) that the Official-Agent Definition requires.
Run scripts/node_shaping_audit.mjs against the assembled spec and fail closed: unresolved cardinality or a leaking legacy route blocks materializing the Agent Node.
Only after pass: true does the daemon proceed to attach a Body, open a transcript, and persist a Work Receipt path for the single archetype selected.
Output Contract
scripts/node_shaping_audit.mjs reads a spec with: workIntent.id, workIntent.signals (coupling, contextPressure, skillBoundary, reviewIndependence, budget, operatorBurden), selectedArchetypes[], and legacyRoutes[] ({ verb, writesIndependentState }). It is safe only when selectedArchetypes has exactly one entry drawn from the seven canonical archetypes and every legacyRoutes[].writesIndependentState is false — an empty legacyRoutes array is safe (vacuously true), but any route proven to write independent state is never safe regardless of how many other routes are clean.
Use scripts/node_shaping_audit.mjs to audit a work-intake spec JSON and return { pass, score, findings, recommendations }.
Anti-Patterns
Zero or Many Archetypes
Novice: Leaves selectedArchetypes empty ("still deciding, will figure it out at runtime") or lists both node and chain to hedge because the signal vector felt ambiguous.
Expert: Forces a single call before any Agent Node materializes. Ambiguity gets resolved in the signal vector, before governance state exists — never by launching more than one archetype and letting them race.
Detection: node_shaping_audit.mjs fires no-archetype-selected (critical) when selectedArchetypes is empty, and multiple-archetypes-selected (critical) when it has more than one entry.
Legacy Verb Smuggles Independent State
Novice: A sortie/spawn/nightshift compatibility shim quietly opens its own session id or starts its own transcript instead of routing through the shared WorkPlan pipeline, producing a second, ungoverned Agent Node for the same WorkIntent.
Expert: Legacy verbs are compatibility metadata only — they annotate provenance ({ sourceVerb: "sortie" }) and then every one of them terminates in the same materialize-one-Agent-Node path.
Detection: node_shaping_audit.mjs fires legacy-route-writes-independent-state (critical) when any legacyRoutes[].writesIndependentState is true.
Invented Archetype
Novice: Introduces an eighth topology ("swarm", "batch", "pipeline") because none of the seven canonical ones felt like a perfect fit, quietly fragmenting a taxonomy the operator was never supposed to see.
Expert: The seven archetypes (node, scout, chain, dag-workgroup, tournament, ambient-watcher, human-gate) are exhaustive by design. A WorkIntent that doesn't fit needs its signal vector rescored against references/seven-archetypes.md, not a new archetype invented.
Detection: node_shaping_audit.mjs fires unknown-archetype (critical) whenever a selected archetype isn't one of the seven canonical names.
References
File
Load When
references/seven-archetypes.md
Need the six-signal definitions or the disambiguation heuristics for picking one of the seven archetypes.
references/legacy-verb-compatibility.md
Need to trace whether a spawn/dispatch/sortie/conjure/nightshift route writes independent state, or need the Official-Agent Definition citation grounding why that matters.
examples/expected-output.md
Need to see a bad intake spec audited, then the same intake fixed and passing.
examples/sample-input.json
Need a complete, already-passing spec to copy as a starting point.
templates/output-template.md
Need a reusable template for the signal vector, archetype decision, and legacy route audit.
schemas/work-intake-spec.schema.json
Need to validate a work-intake JSON payload's structure before auditing it.
scripts/node_shaping_audit.mjs
Need deterministic, fail-closed scoring of a WorkIntent's archetype cardinality and legacy-route safety.
agents/openai.yaml
Need a subagent descriptor for delegated work-intake shaping.
Skill Bundle Index
Every file in this skill, and when to open it. Auto-generated; run scripts/index_references.py --fix.
root
CHANGELOG.md — Work Intake Node Shaping — Changelog — - Initial skill creation - Core process defined: signal-vector-to-archetype mapping plus legacy-verb compatibility audit - Reference files a
README.md — Work Intake Node Shaping — Route one operator WorkIntent to exactly one topology archetype (node, scout, chain, dag-workgroup, tournament, ambient-watcher, human-gate)
examples/expected-output.md — Example Output: Work Intake Node Shaping — Scenario: a nightshift compatibility shim resolves one WorkIntent to two archetypes at once ("dag-workgroup" and an invented "swarm") beca
references/legacy-verb-compatibility.md — Legacy Verb Compatibility — Use this when auditing a spawn/dispatch/sortie/conjure/nightshift code path, or when reviewing a control-plane PR that adds a new
references/seven-archetypes.md — The Seven Topology Archetypes — Use this when scoring a WorkIntent's signal vector and you're not sure which of the seven canonical archetypes it resolves to, or when a pro