WORK-003: work-view post button dead — must do the commit-then-post-to-track two-step

Now: DONT

Input

Context

2026-07-18 live (be tip ba15c38d, work view launched from work/WORK-002): gritzko pressed [post] on the //WORK-002 row and got err: loop: verb 'post' has no plain-args handler //WORK-002: work — the //WORK-002: work tail is the status-bar message+address glue noted cosmetic in WORK-002.

The button mints post '<ticket title>' as an O-invite: views/work/work.js:390-391, SPELL.mintOspell(ctx, "post '" + title + "'").
The throw is core/loop.js:415: plain dispatch demands conv.jab === "args"; core/registry.js:16 — an unmarked module has no plain-args handler. CLI jab post works, so post likely never got the JAB-004 plain-args conversion and the pager spell path is the first caller to notice. Hypothesis — verify.
2026-07-18 later, same session: gritzko clicked the //WORK-003 row itself and got err: NAVNONE: no worktree //WORK-003 //WORK-002: work — work/WORK-003 EXISTED by then (fresh clone in progress/complete), so either the view/nav doesn't see worktrees born after launch, or //WORK-003 anchoring fails some other way; second defect, root-cause separately from the post dispatch.
Same session: gritzko ENTERED the SUBS-053 tree (nav) and typed :get → err: loop: verb 'get' has no plain-args handler — so get dies the same way when run from an OLD tree; hypothesis: registry.resolveVerb's cwd-climb (BE-029, nested jsrc shard wins) resolves the verb module from the NAVIGATED tree's shard, so a stale clone's unconverted verbs shadow the current ones; if true, the whole "no plain-args handler" family is one resolution-scope defect, not per-verb misses.

Goals

RULING (gritzko, chat 2026-07-18): [post] is a TWO-step: (1) commit the wt locally, :post "ticket title"; (2) then :post to the tracked tree (///be).

Both steps run in the ROW's context per WORK-002 — the mutation targets the invite's tree, never the launch tree.
Step 2 conflict handling per POST-032: weave, mark, ONE con row, plain-words state — never hard-err.
Red-first pty repro in the WORK-002 harness shape (test/work/getctx).

Constraints

The WORK-002 fix (get.js targetWt, loop.js NAVNONE mutation gate) is UNLANDED in work/WORK-002 — post's context anchoring needs the same treatment; reference it read-only, expect a weave when both land.
Rulings in force: verbs resolve args via resolve(context,base,rel), pager arg-blind; //authority from arg0 only; errors in plain words; all URI/hash resolution via be/core/resolve_hash.js.

WIP

Design decisions

RULING (gritzko, chat 2026-07-18, supersedes the one-button two-step): --land is VETOED. Bare get / bare post correspond 1:1 to the behind/ahead states, so the ahbeh indicator BECOMES the buttons: [+N] mints bare post (track advance), [-N] mints bare get; no ahbeh — no button. The [post] button does ONLY a commit with the ticket's title (post '<title>').
Split out 2026-07-19 (gritzko): the ahbeh-button UI (bracketed padded [+N]/[-N], [get] retired, colors, view-test amendment) is WORK-004; WORK-003 keeps the post-dispatch defect and the row-context anchoring.

TODOs

pty repro red: click [post] on a wt row → today's no-plain-args-handler err
root-cause: convert post to the plain-args registry shape (or route the spell right)
the two-step: local commit, then post to the tracked tree, both anchored on the row's tree
green: click [post] → commit row + track post row in the ROW's wtlog; launch tree byte-identical

Blockers and bummers

Overlaps unlanded work/WORK-002 in the core/loop.js dispatch area and the target-anchoring idiom.