@ikunin/sprintpilot 1.0.3 → 1.0.5

package/_Sprintpilot/Sprintpilot.md

@@ -25,6 +25,22 @@ When Sprintpilot or the git addon is active:
 - Each story gets its own isolated worktree and branch (`story/<key>`)
 - Commits use conventional format: `feat(<epic>): <title> (<key>)`
 
+### Autopilot configuration
+
+Edit `_Sprintpilot/modules/autopilot/config.yaml`:
+
+| Setting | Default | Values | Purpose |
+|---------|---------|--------|---------|
+| `autopilot.session_story_limit` | `3` | integer ≥ 0 | Stories fully implemented per autopilot run before checkpoint. `0` = unlimited. |
+| `autopilot.retrospective_mode` | `auto` | `auto` / `stop` / `skip` | How epic-end retrospectives are handled (see below). |
+
+`retrospective_mode` options:
+- **`auto`** *(default)* — autopilot writes a deterministic retrospective artifact from `sprint-status.yaml` + `decision-log.yaml`, then continues. Single pass, no external skill call, safe under every CLI.
+- **`stop`** — autopilot pauses at epic completion. Run `/bmad-retrospective` interactively, then re-run `/sprint-autopilot-on` to resume. Use this when you want the full multi-persona discussion as part of your process.
+- **`skip`** — no retrospective artifact is written. **Not recommended** — you lose the epic-level learning record.
+
+Both settings are prompted during `sprintpilot install` (interactive mode) with existing values as defaults, so reinstalls preserve your choices.
+
 ---
 
 ## Full skill reference by lifecycle phase
@@ -59,7 +75,7 @@ See **Mandatory sequence per story** section below.
 
 | Skill | When to use |
 |-------|-------------|
-| `bmad-retrospective` | Run after all stories in an epic are `done`; saves lessons, marks epic `done` |
+| `bmad-retrospective` | Run after all stories in an epic are `done`; saves lessons, marks epic `done`. Under autopilot this is driven by `autopilot.retrospective_mode` (`auto` inline, `stop` to pause for interactive use, or `skip`). |
 
 ---
 

package/_Sprintpilot/manifest.yaml

@@ -1,6 +1,6 @@
 addon:
   name: sprintpilot
-  version: 1.0.3
+  version: 1.0.5
   description: Sprintpilot — autopilot and multi-agent addon for BMad Method (git workflow, parallel agents, autonomous story execution)
   bmad_compatibility: ">=6.2.0"
   modules:

package/_Sprintpilot/modules/autopilot/config.yaml

@@ -18,3 +18,21 @@ autopilot:
   #
   # Set 0 to disable the limit and run until the sprint is complete.
   session_story_limit: 3
+
+  # Retrospective handling at epic completion.
+  #
+  # - "auto" (default): autopilot generates a deterministic retrospective
+  #   from sprint-status.yaml + decision-log.yaml, then continues with the
+  #   next epic. Safe under every CLI including Gemini CLI — no external
+  #   skill call, no persona simulation, single pass.
+  # - "stop": autopilot pauses at epic completion so you can run
+  #   /bmad-retrospective interactively. Re-run /sprint-autopilot-on to
+  #   resume with the next epic.
+  # - "skip": autopilot skips the retrospective entirely and continues.
+  #   NOT RECOMMENDED — you lose the epic-level learning record.
+  #
+  # Background: the external bmad-retrospective skill can enter a
+  # multi-persona discussion loop when driven by autopilot on some CLIs,
+  # which exhausts the token/memory budget. That's why "auto" generates
+  # the artifact inline instead of invoking the external skill.
+  retrospective_mode: auto