sumsub/agent-skills / skills/sumsub-create-workflow
sumsub-create-workflow
VerifiedCreate a Sumsub applicant workflow — verification routing (levels, branches, country/age rules, final-rejection or manual-review steps) or an action workflow (post-verification tagging, notes, source-key changes). TRIGGER on requests to create/build/add a workflow or a flowchart-style routing description. Builds, validates, and saves a draft, then publishes it — gated by live traffic — automatically when the workflow's traffic is low, only after explicit confirmation when it's high. Publishing changes LIVE traffic — see the Danger section. SKIP for editing arbitrary fields of an existing revision, or for questionnaires/levels (separate skills).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 27.6 KB |
| examples/action-workflow.json | config | 2.0 KB |
| examples/aml-routing.json | config | 3.2 KB |
| examples/country-routing.json | config | 1.9 KB |
| examples/risk-routing.json | config | 1.2 KB |
| references/workflow-expressions.md | markdown | 321.2 KB |
| references/workflow-notices.md | markdown | 19.5 KB |
| references/workflow-patterns.md | markdown | 15.9 KB |
| references/workflow-schema.md | markdown | 16.2 KB |
| scripts/build_workflow.pyexecutable | script | 19.3 KB |
| scripts/post_workflow.shexecutable | script | 2.0 KB |
| scripts/publish_workflow.shexecutable | script | 2.2 KB |
| scripts/workflow_api.pyexecutable | script | 14.4 KB |
references/workflow-notices.md
View on GitHub# Workflow validation notices
*(reference — the meaning of each validation notice the API can return. Keys are authoritative
as the API returns them in `notices[].message`; the meanings here are curated to match. The API
is the source of truth — if a notice comes back that isn't listed here, treat its own message as
canonical.)*
`workflow_api.py validate` (and `post`) returns a `notices[]` array. Each notice carries a
`message` (a key in the table below), a `severity`, the offending `nodeId`/`edgeId`, and a
`params` map filling in the specifics — so the notice itself tells you *what*, *where*, and *how
serious*; this table tells you only what the key **means**. **Severity is a property of the
returned notice, not of the message key** (the same key can come back at a different severity),
so read it off the notice — the catalog isn't organized by it.
What a notice's severity means for you:
- **`error`** blocks publish (the API refuses a revision carrying any error notice), so a draft
with open errors is **unfinished**. An error is one of two kinds: a **missing reference** — a
level / tag / source-key / expression path / transition target that doesn't exist — which you
*ground* (reuse, create, or mark pending; never repoint just to go green); or a **structural**
problem in the graph, which you fix in the spec (most of these the builder rejects before
validate even runs). The description tells you which.
- **`warning`** never blocks, but usually marks a real bug — most often a condition reading data
no preceding level produces (a placement issue; see *Place each gate at the earliest point* in
[workflow-patterns.md](workflow-patterns.md)). Read it; don't paper over it.
- **`info`** is advisory.
## Catalog
| key | meaning |
|---------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| `actionLevelsInActionWorkflowsOnly` | An action level is used inside a default (non-action) workflow, where action levels are not allowed. |
| `actionLevelStartRequired` | An action level has incoming edges but no action-level predecessor, so the workflow does not begin with an action level node as required. |
| `approachingActionLimit` | The number of actions on a single node is nearing the per-node maximum. |
| `approachingCriteriaPerEdgeLimit` | The number of condition criteria on a single edge is nearing the per-edge maximum. |
| `approachingEdgeLimit` | The total number of edges in the workflow is nearing the maximum allowed. |
| `approachingNodeLimit` | The total number of nodes in the workflow is nearing the maximum allowed. |
| `approachingReviewButtonIdsLimit` | The number of rejection tags on a final-rejection node is nearing the maximum allowed. |
| `approachingReviewRejectLabelsLimit` | The number of rejection labels on a final-rejection node is nearing the maximum allowed. |
| `approachingTagsLimit` | The number of tags added by a single tags action is nearing the maximum allowed. |
| `approachingTagsToRemoveLimit` | The number of tags removed by a single tags action is nearing the maximum allowed. |
| `approachingTotalActionsLimit` | The total number of actions across all nodes is nearing the maximum allowed for the workflow. |
| `approachingTotalCriteriaLimit` | The total number of condition criteria across all edges is nearing the maximum allowed for the workflow. |
| `approachingTotalTagsLimit` | The total number of tag entries (added and removed) across the workflow is nearing the maximum allowed. |
| `argumentNotFoundInContext` | A function argument references an expression that is not available in the evaluation context. |
| `argumentTypeMismatch` | A function argument resolves to a type that does not match the function's expected parameter type. |
| `blueprintRequired` | A KYT case-creation action's node has no valid associated blueprint. |
| `clientListsBothSidesNotSupported` | A list condition compares two client-list expressions against each other, which is not supported. |
| `conditionalNodeNoConditions` | A branching (exclusive-choice) node has no outgoing edge carrying a condition or review decision, so it cannot branch. |
| `conditionValueAsExpression` | A criterion's literal value looks like a field path, which usually means a Field selector was intended where a plain Value was entered. |
| `consecutiveActionNodes` | An action node's only incoming edge originates from another action node, producing two action nodes back to back. |
| `defaultEdgeMustBeLast` | On a branching (exclusive-choice) node, the default "else" edge — the one with no conditions and no review decisions — is not the last outgoing edge. |
| `deviceCheckWarning` | An expression reads device-intelligence data, but none of the previous levels runs a device check, so that data is never populated for this expression. |
| `disallowedNodeSequence` | A node's type is not permitted to follow the type of the node that precedes it. |
| `duplicateLevel` | The same level is referenced by more than one node within a single workflow. |
| `edgeTargetRequired` | A graph edge points to no target node, leaving that branch of the flow disconnected. |
| `emptyFieldNoActionLevel` | An action field is referenced although no action level precedes this step, so the field will always be empty at evaluation time. |
| `expressionCountryListRecommendation` | A country-field condition enumerates a large number of countries inline; using a client list with alpha-3 country codes is recommended for maintainability (build it with `sumsub-manage-client-lists`). |
| `expressionEmptyList` | An in / not-in condition is checked against an empty list literal, so it can never match anything. |
| `fieldPredefinedValuesWarning` | A criterion compares a vocabulary-backed field against literal values that are not part of that field's predefined vocabulary, so the comparison is unlikely to ever match. |
| `functionArgumentCountMismatch` | A function call supplies a different number of arguments than the function definition requires. |
| `functionArgumentsEmpty` | A function-call criterion was supplied with no arguments at all. |
| `functionNotFound` | A function-call criterion references a function name that is not defined in the workflow's function set. |
| `idDocWarning` | An expression reads data from a specific ID-document set type (such as a questionnaire or a face check), but no preceding node in the flow collects that document type, so the field will be empty when the condition is evaluated. |
| `inEdgeRequired` | A node has no incoming edge and is therefore unreachable in the flow. |
| `invalidExpression` | A criterion uses an expression that is not present in the available data model — typically an unknown or misspelled field path. |
| `invalidLevelTypeChange` | A transition connects levels with incompatible applicant types, for example moving from a KYC level to a KYB level. |
| `invalidSourceKey` | A source-key action references a source key value that does not exist. |
| `kybConversionError` | A KYB 2.0 (WebSDK-next) company level transitions to an old-structure KYB level; the two generations cannot be mixed in one flow. |
| `labelsOrRejectionTagsRequired` | A final-rejection node defines neither rejection labels nor rejection tags, so it has no rejection outcome to apply. |
| `levelApplicantTypeMismatch` | In an action workflow, the target level in the standard workflow has a different applicant type than the source level. |
| `levelFailedInSandbox` | A level node references a production-only level, which will fail when the workflow runs in a sandbox environment. |
| `levelNameEmpty` | An applicant-level node has no level name assigned. |
| `levelNotFound` | A level node references a level that does not exist among the configured levels. |
| `levelReadOnly` | A level node references an inactive, read-only level, which may not be allowed depending on the client's read-only-level limitations. |
| `levelTypeMismatch` | A level node references a level whose type is neither standalone nor actions, and so cannot be used inside a workflow. |
| `missingLevelsForTransition` | An action workflow references levels for its transitions that do not exist in the default (standard) workflow it transitions to. |
| `multipleOutEdgesNotSupported` | A node type that permits only a single outgoing edge has more than one. |
| `nodeHasNoActions` | An action node carries an empty action list and therefore performs no action. |
| `nodeLevelMissingInStandard` | An action transition targets a level that is absent from the standard workflow. |
| `nodeTypeNoConditions` | A node type that does not support conditions has an outgoing edge that carries one. |
| `notesApplicantActionTargetInActionWorkflowsOnly` | A notes action with an applicant-action target is used in a default (non-action) workflow. |
| `outEdgeRequired` | A branching (exclusive-choice) node has no outgoing edges. |
| `questionnaireMissing` | An action-questionnaire expression references a questionnaire that none of the preceding levels collects, so the referenced answers will never be present. |
| `standardLevelsInStandardWorkflowsOnly` | A standalone applicant level is used inside an action workflow, where only action levels are allowed. |
| `standardWorkflowDoesNotExist` | An action transition references a level, but no published standard workflow exists to validate the transition against. |
| `syntaxError` | A condition failed to compile; the underlying compiler error message is included in the notice. |
| `tagDoesNotExist` | A tags action references a tag that does not exist. |
| `tagListEmpty` | A tags action defines neither tags to add nor tags to remove. |
| `tagsApplicantActionTargetInActionWorkflowsOnly` | A tags action with an applicant-action target is used in a default (non-action) workflow. |
| `terminalNodeOutEdgesNotAllowed` | A terminal node (such as an applicant transition) has outgoing edges, which terminal nodes do not allow. |
| `unknownRejectionTag` | A final-rejection node references a review button (rejection tag) that does not exist. |
| `unsupportedProperty` | A deprecated property is used in an expression; a supported replacement property is suggested in its place. |
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |