SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsdi-agent-parameter-sets

di-agent-parameter-sets

Agents & models

View repository

Updated 10-16 05:56

About this skill

Parameter Sets

Never include any UUID in a reply. Project IDs, asset IDs, flow IDs, connection IDs, parameter set IDs — none of these may appear in your response, regardless of whether they came from a tool result or from your own session context. In particular, never open a reply with a phrase like "in your project <uuid>" or "in project 4c2d…". Always identify assets by name only. The only exception is a flow_link field returned ready-made by a tool.

StreamSets only supports string parameters. If the user's request involves a StreamSets flow and any requested parameter is not string type, stop immediately — before calling any tool. Explain the constraint, propose a corrected string-only plan, and wait for confirmation before proceeding. This check must happen before list_streamsets_flows, before create_parameter_set, and before any other tool call.

Parameter sets are named collections of typed parameters stored as project-level assets. They let you define values once and reuse them across multiple flows — switching between environments (dev, test, prod) at job runtime without modifying the flow.

Always Determine Flow Type First

Before calling any tool, determine whether the target flow is:

  • DataStage (batch) — supports all 11 parameter types, value sets, PROJDEF, local parameters
  • StreamSets (streaming) — supports string parameters only; no value sets, no PROJDEF

Most parameter-set features are engine-specific. Getting this wrong leads to silent failures. → See resources/streaming_limitations.md for the full constraints and validation checklist.

What Feature Do You Need?

User need Feature to use
Reuse values across multiple flows Parameter set
Switch between dev / test / prod environments Value set on a parameter set
One-off parameter for a single flow run Local parameter (DataStage only) — use add_local_parameter
Project-level DataStage environment variables PROJDEF
Parameterise a StreamSets flow String-only parameter set
Manage connection values across environments Attach parameter set to connection

Common Tasks

List or find parameter sets

list_parameter_sets(project_id=..., name_filter=...)  # use name_filter to narrow
get_parameter_set(parameter_set_id=..., project_id=...)  # full details incl. value sets

Report by parameter set name only.

Create a reusable parameter set

If the target is a StreamSets flow, verify every parameter type is string before calling any tool — see "Always Determine Flow Type First" above.

create_parameter_set(project_id=..., name=..., parameters=[...])

Confirm by parameter set name and parameters only.

For DataStage flows, all 11 types are supported. → resources/parameter_types.md for the full type table.

Add or modify parameters on an existing set

parameters is a full replacement list — any parameter omitted is deleted. The tool enforces this automatically: if your submitted list would remove existing parameters it returns an error naming the parameters that would be lost and asks you to either include them or pass confirm_replacement=True.

Confirm changes by parameter set name and the updated parameter list.

The safest workflow is still to fetch first and merge:

get_parameter_set(...)               # 1. fetch current parameters
update_parameter_set(                # 2. submit full replacement list
    parameter_set_id=...,
    project_id=...,
    parameters=[...merged...],
)

To intentionally drop parameters, pass confirm_replacement=True:

update_parameter_set(
    parameter_set_id=...,
    project_id=...,
    parameters=[...reduced list...],
    confirm_replacement=True,
)

Create dev / test / prod environments

Value sets let a single parameter set carry multiple environment snapshots:

manage_value_set(action="add",     value_set_name="prod", values={...})
manage_value_set(action="replace", value_set_name="prod", values={...})  # overwrite
manage_value_set(action="remove",  value_set_name="dev")

Confirm by parameter set name and value set name.

Never echo secret values back to the user. When a parameter set contains encrypted parameters, confirm the operation by naming the parameters that were set (e.g. "DB_PASS was stored") — never repeat the actual values in your response. This applies to confirmations, summaries, and plan previews.

→ resources/value_sets.md for full workflow and runtime selection.

Attach a parameter set to a flow

attach_parameter_set_to_flow(project_id=..., flow_id=..., engine=..., parameter_set_name=...)

The registration is saved immediately. After confirming success, always tell the user the next step: to activate the parameters, add references to stage expressions using the correct syntax for the engine:

  • DataStage: #setName.paramName#
  • StreamSets: ${setName__paramName}

Do this in every attach confirmation reply — do not omit the next-step guidance even when the user only asked to attach. Confirm by flow name and parameter set name. → resources/flow_integration.md for reference syntax and workflow.

Attach a parameter set to a connection

attach_parameter_set_to_connection(
    project_id=...,
    connection_id=...,
    parameter_set_name=...,
    property_mappings={"host": "DB_HOST", "port": "DB_PORT", "database": "DB_NAME"},
)

Confirm by connection name and parameter set name. If the tool result includes a runtime value-set selection example, relay it in prose (e.g. "To switch environments at runtime, pass the value set name in the job's runtime parameters") — do not reproduce it as a code block.

Replaces the listed connection property values with #paramSetName.paramName# references and enables the "Use external Parameter Sets" toggle on the connection. Use list_connections to find the connection_id, inspect_project_asset with asset_type="connection" to see the available property names, and list_parameter_sets / get_parameter_set to confirm parameter names. → resources/connection_integration.md for the full workflow.

Audit what parameters a flow uses

get_flow_parameter_references requires a flow_id — never call it with an empty string. When the user gives a flow name, always call list_datastage_flows first to resolve the ID, then call get_flow_parameter_references with it. Do not attempt get_flow_parameter_references before the list call completes.

list_datastage_flows(entity_name=..., project_id=...)   # 1. always do this first
get_flow_parameter_references(flow_id=..., project_id=...)  # 2. then call with the id

Returns external_paramsets (parameter sets registered on the flow) and local_parameters (flow-scoped parameters). Report by flow name, parameter set names, and local parameter names. Use this before deleting a parameter set to confirm the flow still references it, or after attach_parameter_set_to_flow to verify the registration was saved.

DataStage only — engine is optional and defaults to "datastage"; pass engine="streamsets" only if you want the explicit not-supported error. This reports what is registered on the flow, not which stage expressions actually use those parameters; to find #setName.paramName# usage, retrieve the flow with retrieve_datastage_flow_code and search the stage property strings.

Add a local parameter to a DataStage flow

Local parameters are flow-scoped variables defined directly on the flow — not shared with other flows and not part of any parameter set. Reference them in stage expressions as #paramName# (no set name prefix).

add_local_parameter(
    flow_id=..., project_id=...,
    engine="datastage",        # streamsets is not supported — returns an error
    name="SCHEMA",
    type="string",             # all 11 DataStage types supported
    value="PUBLIC",            # optional default
)

Confirm by flow name and parameter name.

  • If a local parameter with the same name already exists it is overwritten (case-sensitive match).
  • type accepts the same aliases as create_parameter_set: integer → int64, float → sfloat, list → enum.
  • enum type requires valid_values=[...].
  • subtype="envvar" is accepted as pass-through but has not been validated against the backend — treat as experimental until confirmed.

Remove a local parameter from a DataStage flow

remove_local_parameter(flow_id=..., project_id=..., engine="datastage", name=...)

Confirm by flow name and parameter name.

Removes the named parameter. Match is case-sensitive. Returns status="removed" on success or status="not_found" if the name does not exist — safe to call without a prior existence check.

Rename or re-describe a set

update_parameter_set(parameter_set_id=..., project_id=..., name="NewName", description="...")

Confirm by old and new names only.

Delete a parameter set

REQUIRED sequence — do not skip any step:

  1. Call list_parameter_sets to resolve the name to an asset ID.
  2. Call get_asset_relationships (asset_type "parameter_set") with that ID before calling delete_parameter_set, regardless of whether you expect references.
  3. If used_by is non-empty, warn the user of every referencing flow or connection and ask for confirmation before proceeding.
  4. If used_by is empty, proceed to delete immediately — no confirmation turn is needed.
  5. Call delete_parameter_set only after step 3 is confirmed (or step 4 applies).
# Step 1
list_parameter_sets(project_id=..., name_filter="MySet")

# Step 2 — always, unconditionally
get_asset_relationships(
    assets=[{"asset_type": "parameter_set", "asset_id": "<resolved_id>"}],
    project_id=...
)

# Step 5 — only after steps 3–4
delete_parameter_set(parameter_set_id=..., project_id=...)

Confirm deletion by parameter set name only.

Permanent and unrecoverable. The tool returns only {'deleted': True, 'parameter_set_id': ...} — it does not check for referencing flows. Deleting a set that is still referenced can leave those flows unable to validate or run until the broken references are removed.

Tool Reference

Tool Purpose
list_parameter_sets List all sets in a project (excludes PROJDEF)
get_parameter_set Full details: parameters + value sets; includes streaming_warnings if the set has StreamSets incompatibilities
create_parameter_set Create a new set
update_parameter_set Rename, re-describe, or replace the parameters list; guards against accidental removal — use confirm_replacement=True to allow drops
delete_parameter_set Permanently delete a set; call get_asset_relationships first to find referencing flows, warn the user, then delete
manage_value_set Add / replace / remove a named value set
attach_parameter_set_to_flow Associate a set with a DataStage or StreamSets flow
attach_parameter_set_to_connection Parameterize connection properties with a parameter set; enables "Use external Parameter Sets" on the connection
get_flow_parameter_references Return external_paramsets and local_parameters registered on a DataStage flow (DataStage only)
add_local_parameter Add or overwrite a local parameter on a DataStage flow (DataStage only)
remove_local_parameter Remove a local parameter from a DataStage flow (DataStage only)

Guardrails

  • See the UUID rule at the top of this skill — never echo any UUID in a reply.
  • PROJDEF is a reserved system name — never create a set with that name.
  • update_parameter_set replaces the entire parameters list. Fetch first, merge, then submit.
  • delete_parameter_set is permanent. Follow the 5-step sequence under "Delete a parameter set" above — get_asset_relationships before delete_parameter_set, unconditionally. Skip the confirmation turn only when used_by is empty; otherwise warn and confirm first.
  • StreamSets: parameter set names must not contain __. Only string parameters are injected.
  • add_local_parameter and remove_local_parameter are DataStage only — passing engine="streamsets" returns an error immediately. On get_flow_parameter_references the engine argument is optional and defaults to "datastage".
  • Call get_flow_parameter_references before add_local_parameter to avoid creating a duplicate under a different case.
  • get_flow_parameter_references reports registration only. A set appearing in external_paramsets does not prove any stage expression uses it, and absence does not prove no #setName.paramName# text remains in the flow.
  • attach_parameter_set_to_connection: both halves of property_mappings are validated before anything is patched — the keys must be properties the connection already has, and the values must be parameters on the named set. Use inspect_project_asset(asset_type="connection") to see available property names. Properties you don't map are left untouched, including secrets.
  • update_datastage_flow replaces the entire flow definition, but it loads the stored flow first, so external_paramsets and local_parameters survive an edit — no re-attach is needed. Confirm with get_flow_parameter_references after edits that matter. → resources/flow_lifecycle_with_parameters.md for the full risk table.
  • When something isn't working as expected → resources/troubleshooting.md.
  • For the most common agent mistakes → resources/common_mistakes.md.