story-kit: create 66_story_pipeline_stage_state_machine_spike

.story_kit/work/1_upcoming/66_story_pipeline_stage_state_machine_spike.md

@@ -0,0 +1,25 @@
+---
+name: Pipeline Stage State Machine Spike
+test_plan: pending
+---
+
+# Story 66: Pipeline Stage State Machine Spike
+
+## User Story
+
+As a Story Kit developer, I want to explore a generalised pipeline stage pattern where each stage is defined by filesystem convention (config.toml, setup script, verify script) and the server drives a uniform loop: enter, validate, setup, execute agent, verify, gate, advance.
+
+## Acceptance Criteria
+
+- [ ] Document the proposed directory structure: .story_kit/pipeline/{stage}/config.toml|setup|verify
+- [ ] Document how work/ directories mirror pipeline/ definitions
+- [ ] Document config.toml schema: agent name, gate policy (advance vs hold)
+- [ ] Document how setup and verify scripts work (executables, exit 0 = pass, env vars passed in)
+- [ ] Document how custom stages would work (user-defined, no Rust changes needed)
+- [ ] Document how human-in-the-loop gates work at each stage (configurable auto-advance vs hold-for-approval)
+- [ ] Evaluate trade-offs vs current hardcoded stage approach
+- [ ] Recommend: implement as-is, simplify, or defer
+
+## Out of Scope
+
+- TBD