"Every muxa reply names the mechanism" is overbroad in CLAUDE.md and decisions.md #46
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
CLAUDE.md's muxa invariants include:
The code is narrower.
{"sent":true}(src/muxa.zig:1210) carries nomechanism; neither does
{"reason":"session_ended","exit_code":N}, northe house
{"error":...,"detail":...}shape. Mechanism accompaniesreplies whose
exit_codeneeds interpreting — which is the ruledocs/decisions.md:3131 is actually arguing for when it says "Only
mechanism == "marks"carries a trustworthy exit code, which is whyevery reply names its mechanism".
So both documents carry the overbroad phrasing and the code is the
narrow one. An agent author reading CLAUDE.md literally would expect a
field that is not there.
Found by the 2026-08-23 drift review. Not fixed there because CLAUDE.md
is the user's instruction file, not something a review pass should edit.
If taken: narrow the wording in both CLAUDE.md and decisions.md:3131 to
scope the rule to replies carrying an exit code.
Migrated from git-collab issue
862ecafdcb519422d337691fa73d49ea35d33fe2.Original author: a73x dev@a73x.sh.
2026-08-29: still true. CLAUDE.md:123 'Every mux a reply names the mechanism'; the code is src/cli/muxa.zig now. Fix is the wording in CLAUDE.md and decisions.md, scoped to replies that carry an exit code.