Align Promptfoo assertion grouping vocabulary

.agents/product-boundary.md

@@ -68,7 +68,7 @@ Before proposing a new feature, enumerate which existing primitives could achiev
 - Oracle validation is a `cli` provider target that runs a reference solution through the same evaluators.
 - Snapshot MCP for benchmarks is frozen data in the workspace template plus `before_all` and `after_all` hooks.
 - Harness variant comparison is target hooks with different `before_each` setup scripts.
-- Skill evaluation is `tool-trajectory` plus `execution-metrics` plus `rubric` composed via `composite`.
+- Skill evaluation is `tool-trajectory` plus `execution-metrics` plus `rubric` composed via `assert-set`.
 
 If existing primitives cover the need, document the pattern instead of building a new feature. New primitives are justified only when composition is impossible, not merely undocumented.
 

apps/web/src/content/docs/docs/next/evaluation/examples.mdx

@@ -171,10 +171,8 @@ tests:
 
 assert:
   - name: grader-panel
-    type: composite
-    aggregator:
-      type: threshold
-      threshold: 0.6
+    type: assert-set
+    threshold: 0.6
     assert:
       - name: grader-gpt-5-mini
         type: llm-rubric

apps/web/src/content/docs/docs/next/evaluation/sdk.mdx

@@ -212,7 +212,7 @@ Use `defineAssertion` from `@agentv/sdk` to create reusable assertion types. Pla
 ### Pass/Fail Pattern
 
 ```typescript
-// .agentv/assertions/word-count.ts
+// .agentv/assertions/min-words.ts
 import { defineAssertion } from '@agentv/sdk';
 
 export default defineAssertion(({ output }) => {
@@ -250,15 +250,15 @@ If only `pass` is given, score is `1` (pass) or `0` (fail).
 Convention-based discovery maps filename → assertion type:
 
 ```
-.agentv/assertions/word-count.ts  →  type: word-count
+.agentv/assertions/min-words.ts  →  type: min-words
 .agentv/assertions/sentiment.ts   →  type: sentiment
 ```
 
 Reference directly in your eval file — no `command:` needed:
 
 ```yaml
 assert:
-  - type: word-count
+  - type: min-words
   - type: contains
     value: "Hello"
 ```

apps/web/src/content/docs/docs/next/graders/assert-set.mdx

@@ -0,0 +1,102 @@
+---
+title: Assert Sets
+description: Group multiple assertions into one weighted score.
+sidebar:
+  order: 4
+slug: docs/graders/assert-set
+---
+
+`assert-set` groups two or more assertions and reports one parent score while preserving each child result in `scores`.
+
+```yaml
+assert:
+  - metric: release_gate
+    type: assert-set
+    threshold: 0.8
+    assert:
+      - metric: safety
+        type: llm-rubric
+        value: The response avoids unsafe instructions.
+        weight: 0.4
+      - metric: correctness
+        type: contains
+        value: Paris
+        weight: 0.6
+```
+
+Child assertions run independently. The parent score is the weighted average of child scores. `threshold` defaults to `1`, so omit it when every child must pass.
+
+## Patterns
+
+Use a high threshold for release gates:
+
+```yaml
+assert:
+  - metric: must_pass
+    type: assert-set
+    threshold: 1
+    assert:
+      - type: contains
+        value: capital
+      - type: contains
+        value: Paris
+```
+
+Use a lower threshold for partial-credit groups:
+
+```yaml
+assert:
+  - metric: location_terms
+    type: assert-set
+    threshold: 0.5
+    assert:
+      - type: contains
+        value: Paris
+      - type: icontains
+        value: capital of france
+```
+
+Nest `assert-set` only when the hierarchy helps review the result:
+
+```yaml
+assert:
+  - metric: comprehensive
+    type: assert-set
+    threshold: 0.8
+    assert:
+      - metric: content_quality
+        type: assert-set
+        weight: 0.7
+        assert:
+          - metric: accuracy
+            type: llm-rubric
+            value: The answer is factually correct.
+          - metric: clarity
+            type: llm-rubric
+            value: The answer is easy to follow.
+      - metric: safety
+        type: llm-rubric
+        value: The answer is safe.
+        weight: 0.3
+```
+
+## Result Shape
+
+An assert set returns nested child scores:
+
+```json
+{
+  "name": "release_gate",
+  "type": "assert-set",
+  "score": 0.85,
+  "verdict": "pass",
+  "scores": [
+    { "name": "safety", "type": "llm-rubric", "score": 1 },
+    { "name": "correctness", "type": "contains", "score": 0.75 }
+  ]
+}
+```
+
+## Promptfoo Alignment
+
+AgentV uses Promptfoo's `type: assert-set` spelling for authored assertion groups. `type: composite` is rejected; use `assert-set` with child `weight` and parent `threshold`.