reducer as wf_mcp capability

docs/core_state_mapping_and_merge.md

@@ -178,7 +178,9 @@ Existing built-in reducers remain distinct:
 
 - `wf.std.replace`
 - `wf.std.append`
+- `wf.std.max`
 - `wf.std.merge_object`
+- `wf.std.set_union`
 
 `wf.std.merge_object` means shallow object merge at the exact destination path, similar
 to `dict.update` or `operator.or_`. It is not a recursive deep merge.
@@ -211,7 +213,6 @@ and belongs in nodes or graph structure.
 Examples a future reducer library could support:
 
 - `max`
-- `set_union`
 - `modulo_add` with configuration such as modulus `10`
 
 ## Implementation Phases

docs/wf_mcp_capability_sources.md

@@ -61,7 +61,8 @@ Expected capabilities:
   `wf.std.truthy`, `wf.std.first_item`, `wf.std.first_item_maybe`,
   `wf.std.first_item_or_none`, `wf.std.last_item`, `wf.std.last_item_or_none`,
   `wf.std.length`, `wf.std.is_empty`.
-- `reducers`: `wf.std.replace`, `wf.std.append`, `wf.std.merge_object`.
+- `reducers`: `wf.std.replace`, `wf.std.append`, `wf.std.max`,
+  `wf.std.merge_object`, `wf.std.set_union`.
 - `prompts`: workflow authoring guide, error-handling guide, mapping guide.
 - `resources`: reference docs for stdlib node behavior.
 

docs/workflow_artifacts.md

@@ -498,6 +498,8 @@ Saved workflow execution eventually needs first-class runtime support for:
 - resume into child run state
 - child final outcome mapping to parent node outcome
 - dependency checks before execution
+- reducer capabilities referenced by declared workflow state fields are saved as
+  direct artifact dependencies just like node specs or tools
 
 The first implementation should prefer artifact validation and dependency
 diagnostics before attempting persistent nested resume.