@delorenj/pjangler 1.2.17 → 1.2.18

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@delorenj/pjangler",
3
- "version": "1.2.17",
3
+ "version": "1.2.18",
4
4
  "description": "Project subsystem bootstrapper CLI",
5
5
  "type": "module",
6
6
  "main": "dist/index.js",
@@ -73,3 +73,35 @@ The loop never ends a pass with work parked waiting on the operator.
73
73
  prerequisite of the dependent and record the rollback.
74
74
 
75
75
  Board status is not proof. Repository evidence and the close gate are proof.
76
+
77
+ ## Post-loop improvement (end-of-batch retro)
78
+
79
+ When the board-clearing loop reaches a resting state — no ready or actionable
80
+ ticket to move and no live worker to monitor (a stop condition above, not a
81
+ mid-flight pause) — the sentinel runs a short end-of-batch retro before the pass
82
+ ends. Because it fires only at rest, it never competes with live ticket flow
83
+ (anti-stall).
84
+
85
+ Answer three questions about the batch just cleared:
86
+
87
+ 1. What hurt this batch?
88
+ 2. What should change?
89
+ 3. Is the fix repo-local or external/template/fleet?
90
+
91
+ ALWAYS record the reflection as a run artifact under
92
+ `_bmad-output/implementation-artifacts/`: the three answers, any repo-local fix
93
+ applied, and a clear description of each proposed external/template/fleet
94
+ improvement. Nothing is silently dropped — the artifact is the durable record.
95
+
96
+ Then act on the answer — do not just note it:
97
+
98
+ - Repo-local, small, safe, and within this pass's scope → implement it now as a
99
+ minor follow-up. If it is larger or riskier, do not start it here; treat it as
100
+ external and surface it below.
101
+ - External / template / fleet (hermes-agent template, fleet config, another repo,
102
+ or an undecided product/architecture decision) → the adapter cannot open issues,
103
+ so surface it loudly instead: post a `tp comment` on the current tracking/active
104
+ issue proposing the follow-up, and flag the operator (the human-facing control
105
+ plane) to file it. Do not attempt to create a ticket through the adapter.
106
+
107
+ The retro is evidence like any other pass output.
@@ -96,5 +96,26 @@ on Plane or Trello.
96
96
  `last_activity_at`, `log_path`).
97
97
  10. Run or follow the session-end ritual; report board status, issues touched,
98
98
  evidence touched, and the active worker issue or blocker.
99
+ 11. **Post-loop improvement (end-of-batch retro).** ONLY when the board-clearing
100
+ loop reaches a resting state this pass — no ready/actionable ticket to move and
101
+ no live worker to monitor (idle, or only backlog / external-blocked work
102
+ remains). This is a retro AT REST: it never runs while a ready ticket or a
103
+ healthy worker exists, and it must not block or delay live ticket flow
104
+ (anti-stall, step 8). Answer three questions about the batch just cleared:
105
+ (1) What hurt this batch? (2) What should change? (3) Is the fix repo-local or
106
+ external/template/fleet? ALWAYS record the reflection as a run artifact
107
+ (evidence) under `_bmad-output/implementation-artifacts/`: the three answers,
108
+ any repo-local fix applied, and a clear description of each proposed
109
+ external/template/fleet improvement. Nothing is silently dropped — the artifact
110
+ is the durable record. Then act on the answer autonomously:
111
+ - **Repo-local** and small, safe, and within this pass's scope: implement it now
112
+ as a minor follow-up inside **{{ target_repo }}**. If it is larger or riskier
113
+ than that, do NOT start it here — treat it as external and surface it.
114
+ - **External / template / fleet** (hermes-agent template, fleet config, another
115
+ repo, or an undecided product/architecture decision): the adapter cannot open
116
+ issues, so surface it loudly instead — post a `tp comment` on the current
117
+ tracking/active issue proposing the follow-up, and flag the operator (the
118
+ human-facing control plane) to file it. Do not attempt to create a ticket
119
+ through the adapter.
99
120
 
100
121
  Do not rely on a hard-coded seed ticket. Query the board every pass.