chore: add measurable impact verification to PR and issue templates

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -70,6 +70,20 @@ body:
     validations:
       required: false
 
+  - type: textarea
+    id: impact
+    attributes:
+      label: Measurable impact of the fix
+      description: |
+        How will we verify the bug is fixed? Provide at least one measurable criterion and a command or method to check it.
+        Example: "Exit code of `bun run cli -- query --question '...'`: before 1 → after 0"
+      placeholder: |
+        | Metric | Before (broken) | After (fixed) | How to measure |
+        |--------|-----------------|---------------|----------------|
+        |        |                 |               |                |
+    validations:
+      required: true
+
   - type: textarea
     id: context
     attributes:

.github/ISSUE_TEMPLATE/feature_request.yml

@@ -41,6 +41,20 @@ body:
     validations:
       required: true
 
+  - type: textarea
+    id: impact
+    attributes:
+      label: Measurable impact
+      description: |
+        How will we know this feature works? List at least one metric with a before/after target and a command or method to measure it.
+        Example: "MCP tool count: before 10 → after 12. Measure: `grep -c createTool src/mcp/tools.ts`"
+      placeholder: |
+        | Metric | Before | After (target) | How to measure |
+        |--------|--------|----------------|----------------|
+        |        |        |                |                |
+    validations:
+      required: true
+
   - type: textarea
     id: alternatives
     attributes:

.github/ISSUE_TEMPLATE/task.yml

@@ -59,6 +59,20 @@ body:
     validations:
       required: false
 
+  - type: textarea
+    id: impact
+    attributes:
+      label: Measurable impact
+      description: |
+        How will we verify the impact of this task? List at least one metric with a before/after target and a reproducible measurement command.
+        Example: "LOC in query-engine.ts: before 1614 → after ≤1200. Measure: `wc -l src/runtime/query-engine.ts`"
+      placeholder: |
+        | Metric | Before | After (target) | How to measure |
+        |--------|--------|----------------|----------------|
+        |        |        |                |                |
+    validations:
+      required: true
+
   - type: dropdown
     id: agent_delegable
     attributes:

.github/pull_request_template.md

@@ -18,6 +18,29 @@
 
 <!-- Key changes, one bullet per logical unit -->
 
+## Impact verification
+
+<!-- Every PR must declare its measurable impact. Fill in at least one row. -->
+
+| Metric | Before | After | How to measure |
+|--------|--------|-------|----------------|
+| <!-- e.g. hard-coded node IDs in query-engine.ts --> | <!-- e.g. 12 --> | <!-- e.g. 0 --> | <!-- e.g. `grep -Ec '[CA]\.' src/runtime/query-engine.ts` --> |
+
+<!--
+Examples of measurable metrics:
+- LOC / module count (refactors): `wc -l src/runtime/query-engine.ts` before and after
+- File count (splits): `ls src/runtime/*.ts | wc -l`
+- Hard-coded IDs removed: `grep -c` specific patterns
+- Test count / coverage: `bun run test` summary
+- CI steps passing: workflow green/red status
+- Build time: `time bun run build`
+- Artifact size: `du -sh .runtime/fpf-index/`
+- New MCP tools: tool count in `src/mcp/tools.ts`
+- Cache hit rate: before/after with `bun run cli -- status`
+
+Pick what fits the PR. The goal: reviewers can independently reproduce the measurement.
+-->
+
 ## Validation
 
 - [ ] `bun run lint` passes locally