Commit 70f05579aa4b3e0a2e99d95cf9ec4e16edfbb553
1 parent
0d602ed4
graphlemon: auto-commit uncommitted worktree changes for 64673231-c2e5-4160-9840-426fae66fe8d
Showing
2 changed files
with
26 additions
and
0 deletions
.claude/agent-memory/planner/MEMORY.md
| 1 | - [loop_round bug already fixed](project_loop_round_bug_already_fixed.md) — off-by-one fix AND the REQ-05/REQ-09 comment above _LOOP_LIMIT are both already present in src/org_loop_route.py; re-verify before replanning either as open | 1 | - [loop_round bug already fixed](project_loop_round_bug_already_fixed.md) — off-by-one fix AND the REQ-05/REQ-09 comment above _LOOP_LIMIT are both already present in src/org_loop_route.py; re-verify before replanning either as open |
| 2 | +- [scheduler.py busy-timeout comment missing](project_scheduler_busy_timeout_comment_missing.md) — as of 2026-08-03 _BUSY_TIMEOUT_MS (scheduler.py:78) has no inline comment; PL does not self-edit, packages as a Unit for Code Agent |
.claude/agent-memory/planner/project_scheduler_busy_timeout_comment_missing.md
0 → 100644
| 1 | +--- | ||
| 2 | +name: project-scheduler-busy-timeout-comment-missing | ||
| 3 | +description: src/scheduler.py's _BUSY_TIMEOUT_MS constant has no comment directly above it as of 2026-08-03; the rationale only lives in the module docstring, far above the constant. | ||
| 4 | +metadata: | ||
| 5 | + type: project | ||
| 6 | +--- | ||
| 7 | + | ||
| 8 | +As of 2026-08-03, `src/scheduler.py` line 78 (`_BUSY_TIMEOUT_MS = 30000`) has no comment | ||
| 9 | +directly above it. The rationale for the 30s value ("concurrent sqlite writers on the shared | ||
| 10 | +db file should retry instead of raising 'database is locked'") exists only inside the module | ||
| 11 | +docstring (lines 21-28), not adjacent to the constant itself. | ||
| 12 | + | ||
| 13 | +**Why:** A user/orchestrator request asked to verify this and, if missing, add exactly one | ||
| 14 | +inline comment line above the constant. This is a source-code edit, which falls outside the | ||
| 15 | +Planning Agent role (SOUL.md: "不写业务代码，不替 Code Agent 实现"). Response was to write a | ||
| 16 | +BRD documenting the gap as a one-line Development Unit for a Code Agent to implement, rather | ||
| 17 | +than editing `src/scheduler.py` directly — even though the change is trivial. | ||
| 18 | + | ||
| 19 | +**How to apply:** If asked again to "check for X comment/fix and edit if missing" as part of a | ||
| 20 | +"write a BRD" request, do the read-only investigation, report the fact, and package any actual | ||
| 21 | +code change as a requirement/acceptance criterion for Code Agent — do not perform the Edit | ||
| 22 | +myself. This mirrors [[project_loop_round_bug_already_fixed]]'s pattern of quick verification | ||
| 23 | +asks embedded in planning requests; the difference here is the fact was "not yet fixed" rather | ||
| 24 | +than "already fixed," so the correct BRD is a genuine (if tiny) actionable requirement instead | ||
| 25 | +of a no-op report. |