diff --git a/.story_kit/work/1_upcoming/95_bug_pipeline_auto_restart_has_no_retry_limit_causing_infinite_loop.md b/.story_kit/work/1_upcoming/95_bug_pipeline_auto_restart_has_no_retry_limit_causing_infinite_loop.md
new file mode 100644
index 0000000..01a0db9
--- /dev/null
+++ b/.story_kit/work/1_upcoming/95_bug_pipeline_auto_restart_has_no_retry_limit_causing_infinite_loop.md
@@ -0,0 +1,28 @@
+---
+name: "Pipeline auto-restart has no retry limit causing infinite loop"
+---
+
+# Bug 95: Pipeline auto-restart has no retry limit causing infinite loop
+
+## Description
+
+When QA (or any agent) fails gates, the pipeline advancement code automatically restarts the agent. There is no retry limit, so if the agent keeps failing gates, it loops forever — spawning, running, failing, restarting. This was observed with story 85 QA which failed gates repeatedly, driving load average to 33 on an M1 Mac.
+
+## How to Reproduce
+
+1. Start an agent on a story where gates will fail (e.g. QA on a story with test failures)\n2. Let the agent complete — server runs gates, gates fail\n3. Observe the server logs — pipeline restarts the agent immediately\n4. Agent fails again, pipeline restarts again, infinite loop
+
+## Actual Result
+
+Agent restarts infinitely with no backoff or retry limit, consuming unbounded CPU and API credits.
+
+## Expected Result
+
+Pipeline advancement should have a max retry count (e.g. 3 attempts). After exhausting retries, mark the story as failed and notify the user instead of restarting.
+
+## Acceptance Criteria
+
+- [ ] Pipeline auto-restart has a configurable max retry count (default 3)
+- [ ] After max retries, agent status is set to failed and no further restarts occur
+- [ ] Server logs clearly indicate when max retries are exhausted
+- [ ] Retry count resets if a human manually restarts the agent