Surface eval outcomes in experiments analyze and align experiment workflows/docs with eval-backed metrics

.github/aw/experiments.md

@@ -135,7 +135,7 @@ Use `${{ experiments.tone }}` tone when writing the issue body.
 
 1. **One dimension** per experiment.
 2. **Falsifiable hypothesis**.
-3. **Primary metric** measurable from workflow run data (artifacts, outputs, duration, tokens).
+3. **Primary metric** measurable from workflow run data (artifacts, outputs, duration, tokens). Prefer `eval:<id>` / `evals.<id>` when success is best measured as a YES/NO question.
 4. **Guardrail metrics** — things that must not degrade. Use `direction: min` + bare number for lower-is-better rates, or `">=0.95"` for higher-is-better.
 5. **Sample size estimate** per variant.
 
@@ -161,6 +161,8 @@ Use `{{#if experiments.prompt_style == "concise" }}` blocks to swap prompt instr
 
 **Typical metrics**: output quality, AI credits, success rate, output length.
 
+When `metric` references `eval:<id>` or `evals.<id>`, declare that eval question under `evals:`. `gh aw experiments analyze` will then show both the metric question text and observed eval YES/NO/UNKNOWN results.
+
 ### Engine & Model
 
 ```yaml
@@ -281,7 +283,7 @@ experiments:
 2. **Instrument** — add `experiments:` and `{{#if experiments.<name> == "<variant>" }}` blocks. Never use `__GH_AW_EXPERIMENTS__*`.
 3. **Compile** — `gh aw compile <workflow-name>`.
 4. **Run** — check activation job step summary for variant assignment.
-5. **Analyse** — once min sample size reached, compare distributions.
+5. **Analyse** — once min sample size reached, compare distributions; for eval-backed metrics, use `gh aw experiments analyze <workflow>` to inspect the resolved question and current eval outcomes.
 6. **Conclude** — rewrite baseline to winning variant, remove `experiments:`, recompile.
 
 ---

.github/workflows/daily-agentrx-trace-optimizer.md

@@ -53,6 +53,10 @@ imports:
       expires: 7d
 features:
   gh-aw-detection: true
+evals:
+  - id: sub_agent_strategy_goal_met
+    question: Does the agent output show that the objective for experiment sub_agent_strategy was successfully completed?
+
 ---
 
 {{#runtime-import? .github/shared-instructions.md}}

.github/workflows/dataflow-pr-discussion-dataset.md

@@ -106,6 +106,10 @@ experiments:
     start_date: "2026-06-05"
     issue: 37102
 strict: true
+evals:
+  - id: caveman_mode_goal_met
+    question: Does the agent output show that the objective for experiment caveman_mode was successfully completed?
+
 ---
 
 {{#if experiments.caveman_mode == 'yes' }}

.github/workflows/deep-report.md

@@ -86,6 +86,10 @@ imports:
 
 
   - shared/otlp.md
+evals:
+  - id: output_format_goal_met
+    question: Does the agent output show that the objective for experiment output_format was successfully completed?
+
 ---
 
 ### DeepReport - Intelligence Gathering Agent

.github/workflows/gpclean.md

@@ -95,6 +95,8 @@ evals:
     question: Does the agent output confirm that go.mod was analyzed for GPL-licensed transitive dependencies?
   - id: decision-explained
     question: Does the agent output include an explanation of why a GPL issue was created or why noop was called?
+  - id: tool_verbosity_goal_met
+    question: Does the agent output show that the objective for experiment tool_verbosity was successfully completed?
 
 ---
 

.github/workflows/smoke-antigravity.md

@@ -72,6 +72,10 @@ safe-outputs:
 timeout-minutes: 10
 features:
   gh-aw-detection: false
+evals:
+  - id: sub_agent_strategy_goal_met
+    question: Does the agent output show that the objective for experiment sub_agent_strategy was successfully completed?
+
 ---
 
 # Smoke Test: Antigravity Engine Validation