Skip to content
Merged
Show file tree
Hide file tree
Changes from 2 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions .agents/skills/issue-implement/SKILL.md
Original file line number Diff line number Diff line change
Expand Up @@ -506,6 +506,8 @@ When in doubt between "non-actionable code feedback" and "non-code conversation"

**A second race on the same structure is a missing primitive, not another window.** When a finding is an interleaving — a read and a later act on state another run, process or reused id can change in between, or two writes a crash or restart can split — a fix that only narrows the gap (a re-read, a tag, an extra check) leaves the next gap for the next round. Whether your last fix closed the class is the next review's call, not yours: a second finding that reopens the same race (the same invariant an earlier round already fixed; a reviewer's "fresh evidence beyond the earlier … fix" says so) is the trigger, even when that fix was a restructure you said closed the class. A separate race on the same structure, such as a delete race after a create race was closed, is not; fix it here. Then name the atomic operation the closing fix needs (a conditional write, a per-attempt id, a namespace, one write or a recorded intent recovery replays), file it or point to the issue that already holds the class, and answer this thread and every later one on that race there. In this PR, fix only the gaps the existing primitives can close.

**From the fourth review pass, count instead of judging.** A chain of different cases on one structure each passes the "separate race" test, so the judgment above never fires on it. On a PR's fourth review pass, and every later one, that has a finding on a structure an earlier pass already changed (the same check, record, list or key, whatever the case), answer the question above on the PR before fixing: name the primitive and file it, or say in one line why these findings share none.
Comment thread
jhoffner marked this conversation as resolved.
Resolved

#### 10.4: Take action and reply

Process each comment in its bucket:
Expand Down
1 change: 1 addition & 0 deletions docs/internal/cycle-ledger.md
Original file line number Diff line number Diff line change
Expand Up @@ -80,3 +80,4 @@ earlier cycle files are records and are not rewritten.
| 25 | one task primitive epic wrap (FIX-2056) (2026-10-24) | [`cycle-25.md`](cycle-ledger/cycle-25.md) |
| 26 | one waiting-on-a-person epic wrap (FIX-2058) (2026-10-27) | [`cycle-26.md`](cycle-ledger/cycle-26.md) |
| 27 | harness subscription-usage epic wrap (FIX-2139) (2026-10-29) | [`cycle-27.md`](cycle-ledger/cycle-27.md) |
| 29 | same pieces, different apps epic wrap (FIX-2152) (2026-10-30); 28 is the open FIX-2071 draft | [`cycle-29.md`](cycle-ledger/cycle-29.md) |
Loading
Loading