Assign Risk Power User Role
Purpose
Assign the Risk Power User role to eligible current default USER owners
configured by Assign Default Process Owners. Do not infer owners and do not
modify process-owner properties.
Consultant-facing behavior
Follow the offering Consultant Interaction Contract. Present access status for eligible owners, exceptions, and completion in compact Markdown tables. Use Person, Affected process, Current result, and Approval or next action when applicable. The consultant confirms a table or chat decision; do not ask them to inspect or approve request bodies, identifiers, polling data, or JSON files.
Usage
Use this skill when planning, sequencing, or validating the Assign Risk Power User Role setup task under Assign Initial Users.
Required upstream skill
Require the verified coverage result from the parent Risk Administrator Assignment Verification. Do not begin Risk Power User discovery, planning, or access-request preparation until the parent reports verified coverage for the selected project and environment.
Core execution delegation
Prepare every Fusion or BOSS request in this skill and defer its execution to
core.fusion-api.rest-api-client; that skill's SKILL.md governs when its
Python client may run. Do not import, launch, or duplicate the Core client's
OAuth, token-cache, transport, retry, or response-processing behavior here.
Use scripts/transform_results.py only after
the relevant Core results have been returned. It performs deterministic joins,
eligibility checks, and postcondition evaluation; it neither creates request
manifests nor calls Fusion. Its machine-readable input and output are internal;
use its Markdown table result when communicating with the consultant.
A successful, valid role response that omits members means no people are
assigned, just like members: []. Report “No people are currently assigned to
the Risk Power User role.” and continue the existing eligibility and approval
flow. An explicitly null or non-list members value remains invalid. Empty
membership does not verify provisioning or automatically complete this task.
Authentication
Use core.fusion-api.rest-api-client for authentication and every Fusion
request. Never request, persist, or display a Fusion password, access token,
refresh token, or client secret. At the start of discovery, prepare this
Core-executed, read-only request to resolve the signed-in user:
GET /hcmRestApi/resources/11.13.18.05/selfDetails?fields=Username
The returned Username is supplied by Fusion from the security token. It is
the only permitted requester identity for this workflow; do not present it for
confirmation or accept an override. Do not decode a token in this skill.
Before supplying current_user to the discovery transformer, accept either a
non-empty root Username or exactly one items entry with a non-empty
Username and hasMore=false. For the collection form, construct the
transformer's minimal root-shaped current_user input in memory. Treat zero or
multiple entries, pagination, malformed data, or a blank username as an
identity blocker. Do not persist or display the signed-in username or the
normalized identity object.
Prepare the matching authenticated-user resolution through:
GET /hcmRestApi/scim/Users?filter=userName eq "<selfDetails Username>"&count=2
Its unique, active, case-insensitive SCIM userName match provides the id
used as requestedByUserGuid. If either authenticated-user result is missing,
ambiguous, inactive, or mismatched, stop before preparing any AAR submission.
Execution
Prepare the signed-in-user request shown above. Normalize its valid root-shaped or collection-shaped result as described above, then derive the requester solely from that result; do not request consultant confirmation or an alternate username.
Prepare these read-only Core requests, using that authenticated username:
GET /api/boss/data/objects/ora/riskMgmtRisk/core/v1/referenceDataValues?$filter=referenceCode!='SOURCE.ORA.ORA' AND referenceCode!='SOLUTION.FND.ORA' AND attributeCode='PROCESSES.ORA'&$fields=valueCode,referenceCode,attributeCode,levelOne,levelTwo,levelThree,description&$sortBy=levelTwo:asc,referenceCode:asc&$limit=100&$offset=0 GET /api/boss/data/objects/ora/riskMgmtRisk/core/v1/referenceDataValueProperties/$views/detailed?$limit=100&$offset=0&$sortBy=valueCode:asc GET /hcmRestApi/scim/Users?filter=userName eq "<selfDetails Username>"&count=2 GET /hcmRestApi/scim/Roles?filter=name eq "ORA_GTG_RISK_POWER_USER_JOB"&count=2Run the transformer with
--stage discovery,process_catalog_endpointset exactly to/api/boss/data/objects/ora/riskMgmtRisk/core/v1/referenceDataValues, and those results passed under their documented response IDs. If it returnsneeds_role_membership, prepareGET /hcmRestApi/scim/Roles/<role_id>. For each value inneeds_user_lookups, prepareGET /hcmRestApi/scim/Users/<user_id>. Re-run discovery after those reads. It separatescandidates_requiring_decision,already_assigned, andineligibleowners. If the first group is empty, finish with No changes required; do not ask for approvals or make further calls. Read each nonempty page whilehasMore=true, advancing$offsetby the returned page size; BooleanhasMore=falsecompletes the inventory. This direct business object returns provided records only. Do not use theallReferenceDataValuesunion extraction, and stop if the declared direct source is not present. Stop if authenticated-user or role resolution is ambiguous.Without prompting the customer, prepare a fresh role-membership read and a fresh SCIM user read for every
candidates_requiring_decisionentry. Run the transformer with--stage preflight. It returns only assignments that remain eligible at this final check. If it returnsawaiting_core_results, perform its requested reads and re-run preflight.During discovery and preflight, immediately perform each required safe read and re-run the applicable transformer stage until it produces an approval table, verified no-change result, or genuine blocker. Do not end a response after an internal read, normalization step, or
awaiting_core_results. In these read-only phases, an otherwise unspecifiedcontinueorresumerequest authorizes the outstanding reads only; it is not a person selection or approval. Once the approval table is ready, stop for the explicit approval.Present the exact environment and target users as an approval table. Ask the customer to select the entries to submit and state clearly that their response is the final approval for the selected people and will immediately submit only those requests; do not ask again. Retain every unselected eligible person as a remaining candidate. A partial approval is never an implicit decision to skip the other people. Retain the following body internally for execution:
POST /fscmRestApi/resources/11.13.18.05/advancedAccessRequests/action/createAccessRequest
Content-Type: application/vnd.oracle.adf.action+json
{
"requestedByUserGuid": "<requested_by_user_guid>",
"requestedForUserGuid": "<user_guid>",
"isTemporary": false,
"justification": "Needed for role as business process owner",
"requestAttributes": [{
"roleCode": "ORA_GTG_RISK_POWER_USER_JOB",
"securityContextId": "",
"securityValueId": "",
"OraGrcAssignAccess": "Y"
}]
}
Access requests are individual, never retried automatically, and never use
bulk or PATCH.
6. Record each returned access-request ID with its user and process details.
Before the first status read, record poll_started_at in UTC. Supply the
positive, caller-approved poll_interval_seconds and
max_poll_wait_seconds values, plus the UTC observed_at timestamp, to
every --stage poll transformer input. The transformer returns
elapsed_wait_seconds and, while work remains, a
next_check_after_seconds value. For each returned
pending_aar_ids, prepare:
GET /fscmRestApi/resources/11.13.18.05/advancedAccessRequestApprovals?q=AccessRequestId=<aar_id>&limit=2
Re-run it only after the returned next-check interval. Never exceed
max_poll_wait_seconds. When the transformer returns timed_out, stop
polling and persist a redacted process_timeout diagnostic containing the
environment, operation, AAR IDs, last known statuses, elapsed wait, and
next-check guidance. Do not submit another request or continue polling
without a fresh customer decision. When needs_role_membership is true,
prepare a fresh GET /hcmRestApi/scim/Roles/<role_id>.
7. Run the transformer with --stage verify and that membership result. Do not
mark a successful access request complete until its user is present in the
verified role membership. Then prepare a fresh role-membership read and a
fresh SCIM user read for every unselected candidate and every person whose
submitted request was not verified. Run the transformer with
--stage continuation, passing the verify executions as
verification_executions and the unselected candidates as
remaining_candidates.
If continuation returns awaiting_remaining_decision, present the remaining
table and ask a direct question that offers two paths: assign access to one
or more listed people, or leave every remaining assignment unassigned and
resume the runbook. Make clear that the second path records this task as
complete with the remaining assignments skipped. Describe the available
response intents—selected people, all remaining people, or moving on—without
requiring literal wording. Each later approval covers only its selected
people and requires the same fresh preflight. Treat an otherwise unspecified
request to continue, a polling response, or a prior partial approval as
ambiguous and re-present the table and question. If the consultant
unambiguously chooses to skip remaining people and proceed, record
consultant_decision: SKIP_REMAINING_AND_PROCEED internally and return its
completed_deferred result to the calling runbook. The caller
then runs its next eligible task, or finishes if none remains. Otherwise,
return completed_verified only for completed_all_resolved or a verified
no-changes-required outcome.
An explicit request to defer and switch tasks records the outstanding people
and returns deferred; it is not a skip decision. A failed, uncertain, or
unverified submission or membership result returns blocked.
Persist the readable receipt and exceptions after every batch.
The run creates a customer-controlled receipt plus redacted receipt, exception, run-record, and token-usage artifacts under the project workspace.
Core result ledger
Use these stable response IDs when passing returned data to the transformer. They are internal workflow labels, not customer-facing content.
| Response ID | Core result |
|---|---|
current_user |
Signed-in user lookup. Its Username is the sole requester identity. |
process_catalog |
Direct BOSS referenceDataValues provided-process catalog; the union extraction is not eligible. |
properties |
BOSS default-owner properties. |
authenticated_user |
Exact active SCIM result for the signed-in user. |
role |
Risk Power User role SCIM result. |
role_members |
Current role membership; read again during preflight, verification, and continuation. |
user:<user_guid> |
A default owner, preflight candidate, or continuation candidate's SCIM result. |
poll:<aar_id> |
One Advanced Access Request status result. |
Execution handoff
Invoke core.fusion-api.rest-api-client for each prepared discovery,
access-request, polling, and verification operation. Keep the returned IDs and
redacted responses in this skill's durable receipt; never run a pillar helper
as an alternate Fusion client.
Preflight and failure handling
Before executing any Core request, verify the selected project workspace,
environment context, required prior owner configuration, request-plan/result
environment match, and the presence of every required Core response. Treat a
missing, inactive, ambiguous, or mismatched signed-in-user result as an
auth_or_identity exception; an ambiguous role as
catalog_or_metadata_mismatch; and a missing
or inactive owner as an actionable eligibility exception. Let the Core skill
handle its documented one-time authentication refresh for read failures.
Do not retry an access-request submission automatically. On a submission,
status, or membership-verification failure, persist redacted diagnostics and
the completed or partial receipt, then require fresh customer approval before
another submission attempt. The receipt distinguishes verified,
remaining_pending_decision, resolved_without_change, and
skipped_by_explicit_consultant_decision; only the final category permits the
runbook to move on while a person remains unassigned.