code_diff.patch
2.39 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
diff --git a/.claude/agent-memory/coder/MEMORY.md b/.claude/agent-memory/coder/MEMORY.md
new file mode 100644
index 0000000..dbfa66d
--- /dev/null
+++ b/.claude/agent-memory/coder/MEMORY.md
@@ -0,0 +1 @@
+- [Loop-limit comment task keeps re-dispatching with zero diff](task_dispatch_loop_limit_comment_repeat.md) — src/org_loop_route.py:38 comment already exists; verify and report, don't re-add
diff --git a/.claude/agent-memory/coder/task_dispatch_loop_limit_comment_repeat.md b/.claude/agent-memory/coder/task_dispatch_loop_limit_comment_repeat.md
new file mode 100644
index 0000000..dc0c60e
--- /dev/null
+++ b/.claude/agent-memory/coder/task_dispatch_loop_limit_comment_repeat.md
@@ -0,0 +1,26 @@
+---
+name: task-dispatch-loop-limit-comment-repeat
+description: The "_LOOP_LIMIT boundary comment missing" task keeps getting re-dispatched even though the comment has been present and correct since it was added; verified 5+ times now with zero diff each time
+metadata:
+ type: project
+---
+
+`src/org_loop_route.py:38` has had the REQ-05/REQ-09 loop_round-boundary comment above
+`_LOOP_LIMIT = 3` since it was added under U-016 (commit `59749fc`). A goal describing this
+exact gap ("no comment there, add one") has now been dispatched to the coder at least 5
+independent times (per the user's global auto-memory index and this session's own SDD, which
+called it the "4th independent verification" before this one, the 5th). Every time, the correct
+action was to verify and make zero changes.
+
+**Why:** the Org Graph / planner pipeline has no terminal task state for "investigated, already
+satisfied, confirmed zero-diff" that's distinct from `done` / `blocked_pending_human` — this was
+flagged as an out-of-scope process gap in this session's SDD and in the referenced auto-memory
+entry. Without that state, tasks whose fix already landed apparently keep getting re-surfaced as
+open work instead of being marked resolved.
+
+**How to apply:** if dispatched this same goal again, re-verify quickly (grep the line, run
+`python3 -m unittest src.test_org_loop_route -v`), confirm zero diff, and report the same finding
+rather than assuming something changed. Do not silently "fix" it by re-adding/rewording the
+comment — that would create unnecessary diff noise. If this keeps recurring, the real fix is
+upstream (a Planner/Org Graph terminal state for zero-diff-confirmed tasks), not another coder
+turn on this file.