feat: add strategy-aware benchmark artifacts

API.md

@@ -733,7 +733,7 @@ interfacectl summarize-generation-session --session-dir <path>
 
 ### `compare-generation-sessions`
 
-Compares one unguided baseline session against one prepared guided session for the same implementation brief.
+Compares two tracked generation sessions for the same implementation brief.
 
 **Synopsis:**
 ```bash
@@ -742,7 +742,7 @@ interfacectl compare-generation-sessions --baseline-session-dir <path> --guided-
 
 **Description:**
 - Requires both sessions to target the same surface, use the same tool, and freeze the same brief file.
-- Requires `guidanceMode=unguided` for the baseline session and `guidanceMode=prepared` for the guided session.
+- Works with any valid guidance-strategy pair; the output records each session’s concrete strategy.
 - Computes first-attempt finding deltas, attempts-to-acceptable-outcome delta, rubric deltas, and goal checks.
 - Writes `comparison.json` and `comparison.md`.
 - Canonical schema lives at `packages/interfacectl-cli/schemas/generation-session-comparison.schema.json`.
@@ -809,7 +809,7 @@ interfacectl summarize-generation-benchmark --comparisons <path[,path...]> [--su
 ```
 
 **Description:**
-- Summarizes whether guided sessions reduced first-attempt blocking findings, reached acceptable outcomes no later, and improved rubric dimensions.
+- Summarizes whether the compared candidate sessions reduced first-attempt blocking findings, reached acceptable outcomes no later, and improved rubric dimensions.
 - Aggregates accepted/rejected/proposed suggestion counts across surfaces.
 - Writes `benchmark-report.json` and `benchmark-report.md`.
 - Canonical schema lives at `packages/interfacectl-cli/schemas/generation-benchmark-report.schema.json`.

README.md

@@ -199,7 +199,7 @@ interfacectl summarize-generation-session --session-dir <path>
 
 ### `compare-generation-sessions`
 
-Compares one unguided baseline session against one guided prepared session for the same brief and writes deterministic comparison artifacts.
+Compares two tracked generation sessions for the same brief and writes deterministic comparison artifacts.
 
 ```bash
 interfacectl compare-generation-sessions --baseline-session-dir <path> --guided-session-dir <path> [--out-dir <path>]

docs/ai-generator-adapter-quickstart.md

@@ -6,12 +6,12 @@ Use this flow when a local agent or hosted generator needs contract-aware guidan
 
 1. Compile the contract into a generation bundle.
 2. For local agents, resolve the bundle into one agent-ready payload with `prepare-generation`.
-3. Freeze one tracked session with `init-generation-session` when you want iteration evidence or a guided-vs-unguided benchmark.
+3. Freeze one tracked session with `init-generation-session` when you want iteration evidence or a strategy benchmark.
 4. Generate or edit UI.
 5. Run `record-generation-attempt` for each attempt.
 6. Optionally run `review-generation-attempt` when a `warn` result is explicitly acceptable.
 7. Run `summarize-generation-session` to aggregate progress.
-8. Use `compare-generation-sessions`, `suggest-contract-deltas`, and `summarize-generation-benchmark` when you are proving guided-vs-unguided outcomes.
+8. Use `compare-generation-sessions`, `suggest-contract-deltas`, and `summarize-generation-benchmark` when you are proving one guidance strategy against another.
 9. Use `validate-generation` directly when you need an ad hoc post-generation check without a tracked session.
 
 ## Step 1: compile the bundle
@@ -82,7 +82,7 @@ interfacectl init-generation-session \
   --bundle-root ./artifacts/generation-bundles/surfaces-web \
   --surface surfaces-web \
   --workspace-root . \
-  --guidance-mode prepared \
+  --guidance-strategy prompt-summary \
   --brief-file ./artifacts/generation-briefs/surfaces-web.md
 
 interfacectl record-generation-attempt \
@@ -100,15 +100,15 @@ interfacectl summarize-generation-session \
 
 This loop freezes the bundle revision, records each assessment, and emits canonical run artifacts for downstream consumers.
 
-For an A/B proof loop, run one session with `--guidance-mode unguided` and the same `--brief-file`, run another with `--guidance-mode prepared`, then compare them:
+For an A/B proof loop, run two sessions with the same `--brief-file`, then compare the strategies you want to evaluate. For example, compare `prompt-summary` against `json-primary`:
 
 ```bash
 interfacectl compare-generation-sessions \
-  --baseline-session-dir ./artifacts/generation-sessions/surfaces-web/baseline-unguided \
-  --guided-session-dir ./artifacts/generation-sessions/surfaces-web/guided-prepared
+  --baseline-session-dir ./artifacts/generation-sessions/surfaces-web/prompt-summary \
+  --guided-session-dir ./artifacts/generation-sessions/surfaces-web/json-primary
 
 interfacectl suggest-contract-deltas \
-  --session-dir ./artifacts/generation-sessions/surfaces-web/guided-prepared
+  --session-dir ./artifacts/generation-sessions/surfaces-web/json-primary
 ```
 
 ## HTTP mode

docs/generator-consumption.md

@@ -17,7 +17,7 @@ For workspace agents:
 
 1. Run `interfacectl compile --contract <path> --out <bundleDir>`.
 2. Run `interfacectl prepare-generation --bundle-root <bundleDir> --surface <id>`.
-3. Optionally run `interfacectl init-generation-session --bundle-root <bundleDir> --surface <id> --workspace-root <path> --guidance-mode prepared --brief-file <path>` when you want tracked iteration evidence or a benchmark-ready guided session.
+3. Optionally run `interfacectl init-generation-session --bundle-root <bundleDir> --surface <id> --workspace-root <path> --guidance-strategy <prompt-summary|json-primary|unguided> --brief-file <path>` when you want tracked iteration evidence or a benchmark-ready session.
 4. Feed the resulting prepared JSON into the agent.
 5. Generate only inside the surface-owned boundary.
 6. Either run `interfacectl validate-generation --mode workspace` directly, or run `interfacectl record-generation-attempt` for a tracked session.
@@ -80,7 +80,7 @@ When you need auditable iteration history, use the canonical session commands ra
 3. `interfacectl review-generation-attempt` when a warning is explicitly acceptable
 4. `interfacectl summarize-generation-session`
 
-For the guided-vs-unguided proof loop, compare two sessions that froze the same `--brief-file`:
+For the strategy-benchmark loop, compare two sessions that froze the same `--brief-file`, such as `prompt-summary` vs `json-primary` or `unguided` vs a guided strategy:
 
 1. `interfacectl compare-generation-sessions`
 2. `interfacectl suggest-contract-deltas`

packages/interfacectl-cli/dist/commands/generation-session.d.ts

@@ -5,9 +5,18 @@ export interface InitGenerationSessionCommandOptions {
     tool?: string;
     sessionId?: string;
     artifactsRoot?: string;
+    guidanceStrategy?: string;
     guidanceMode?: string;
     briefFile?: string;
 }
+export interface PrepareGenerationHandoffCommandOptions {
+    sessionDir?: string;
+    guidanceStrategy?: string;
+    acceptedSuggestionsFile?: string;
+    designerNotesFile?: string;
+    findingCodes?: string;
+    outPath?: string;
+}
 export interface RecordGenerationAttemptCommandOptions {
     sessionDir?: string;
     assessmentFile?: string;
@@ -47,6 +56,7 @@ export interface SummarizeGenerationBenchmarkCommandOptions {
     outDir?: string;
 }
 export declare function runInitGenerationSessionCommand(options: InitGenerationSessionCommandOptions): Promise<number>;
+export declare function runPrepareGenerationHandoffCommand(options: PrepareGenerationHandoffCommandOptions): Promise<number>;
 export declare function runRecordGenerationAttemptCommand(options: RecordGenerationAttemptCommandOptions): Promise<number>;
 export declare function runCaptureGenerationPreviewCommand(options: CaptureGenerationPreviewCommandOptions): Promise<number>;
 export declare function runReviewGenerationAttemptCommand(options: ReviewGenerationAttemptCommandOptions): Promise<number>;