refactor(host)!: retire the skill.invoke RPC for the gesture boundary

Invocation is an ordinary session.prompt again: the pre-step gesture
boundary makes it deterministic host-side for every front end, so the
dedicated RPC (handler, wire schema, error codes, client face, fixtures)
and ui-skill's claim machinery are net deletions. The menu keeps decision
21 exactly — a pick lands literal /name text — plus the user-only marker
from skill.list's modelInvocable flag.

packages/client/ui-skill/README.md

@@ -4,7 +4,7 @@ English | [中文](README.zh.md)
 
 Skill invocation source, browser half: registers the `/`-trigger `skill` source into `ctx.slash`. Ordinary-session candidates come from the `skill.list` RPC addressed by the per-call `ClientSessionContext` projection's `{sessionId}`, with the host resolving `cwd` from the session header. The host serves every user-invocable skill; a `modelInvocable: false` entry (a `disable-model-invocation` skill, whose only entry point is this path) wears the user-only marker as a description prefix in the active language. Catalog-addressed continuable children resolve no skill candidates locally because the existing skill RPC requires an attached session; viewing their persisted history must not activate them. Catalogs cache per ordinary session with a single-flight fetch; the scope-birth `warm` hook prewarms the session's entry and `connection/reset` clears everything. Results filter by `startsWith(query)`.
 
-A menu pick or an entered `/name [args]` line claims the composer into an args-tolerant `skill.invoke` transaction (`matchEnter` strong-waits the catalog; an unknown name answers undefined and stays a plain prompt). A skill name shared with a host command resolves to the command: adjudication polls sources in registration order and the web bundle mounts ui-command ahead of this source — deliberate precedence, matching peer products. Submit trims the args, keeps blank args off the wire, and folds an RPC refusal into the composer's error outcome; the host renders the skill body and injects it as a user message before starting the turn, so invocation is deterministic for every user-invocable skill. The RPC rides the plugin's root-context connection captured at registration — the source never reads services off a per-call argument. Draft chip visuals still derive from the `lexicon` scan; the legacy `<skill>name</skill>` reference codec is gone (decision 21 removal cut) and `matchSpace` stays unimplemented — menu and enter own the skill flows.
+A pick lands the literal `/name ` text and the prompt ships the same literal (decision 21) — this source implements no adjudication hooks and no reference codec (the legacy `<skill>name</skill>` form is gone with the removal cut). Determinism lives host-side: the pre-step gesture boundary (`dsh-tool-skill`) recognizes whitespace-bounded `/name` tokens naming user-invocable skills anywhere in a user message and injects the rendered `<skill_content>` for every front end, so a menu pick, a hand-typed token, and a TUI/ACP prompt all load the skill the same way. A name shared with a host command still resolves to the command: adjudication claims the line client-side before it ever becomes a prompt — deliberate precedence, matching peer products. The list RPC rides the plugin's root-context connection captured at registration — the source never reads services off a per-call argument; draft chip visuals derive from the `lexicon` scan.
 
 A failed `skill.list` throws from `candidates`, which the slash shell logs and folds into a silent menu-group drop — the menu shows only pending/ready states.
 
@@ -20,11 +20,11 @@ The browser plugin also registers a keyed `skill` toolview in `conversation.chat
 
 #### What the model sees
 
-A claimed invocation never ships the `/name` literal. The host (`skill.invoke`) renders the canonical `<skill_content>` block — the same `renderSkillContent` output the `skill` tool returns — appends the user's trailing text after a blank line, and injects the whole as one user-role message carrying the `skill-invocation` source, immediately starting a turn. Loading is deterministic: the model receives the full body without being asked to call the `skill` tool, and the catalog (rendered by `dsh-tool-skill`) tells it not to re-load an inline-injected skill.
+The user's message reaches the model verbatim, `/name` literal included. The host's pre-step boundary (`dsh-tool-skill`) then appends the canonical `<skill_content>` block — the same `renderSkillContent` output the `skill` tool returns — as injected instructions context at the end of that step's injections, closest to the model's answer. Loading is deterministic: the model receives the full body without being asked to call the `skill` tool, and the catalog tells it not to re-load an inline-injected skill.
 
 #### Token effect
 
-One invocation adds the rendered skill body plus the trailing text to that turn's user message — the same cost as the model loading the skill through the tool, paid unconditionally instead of at the model's discretion. Menu browsing and the candidate fetch add zero model tokens.
+One invocation adds the rendered skill body to that turn as injected context — the same cost as the model loading the skill through the tool, paid unconditionally instead of at the model's discretion. Menu browsing and the candidate fetch add zero model tokens.
 
 #### KV Cache effect
 
@@ -33,5 +33,5 @@ Append-only: the injected message lands after the reusable history prefix. This
 ## Known Limitations and Deferred Work
 
 - **Result-only history pages use the generic row** — keyed dispatch needs the paired call in the runtime window; pagination that leaves the call outside has no tool identity. This client presentation feature does not extend the history wire contract to recover it.
-- **Enter waits on the catalog once** — `matchEnter` strong-waits the session's first catalog fetch before answering, so an enter racing a cold cache resolves against the settled catalog rather than silently missing. A menu opened before the prewarm settles still shows no skill candidates for that keystroke.
-- **Text is the truth** — the reference is plain draft text; a hand-typed identical token is the same reference. Chip visuals derive from the lexicon scan; no occurrence identity or position tracking (componentized chips are a ledger item).
+- **Text is the truth** — the reference is plain draft text; a hand-typed identical token is the same reference, and the host gesture boundary judges the sent text, not the menu interaction. Chip visuals derive from the lexicon scan; no occurrence identity, position tracking, or structured reference payload on the prompt wire (both are ledger items).
+- **A menu opened before the prewarm settles** shows no skill candidates for that keystroke; the next keystroke re-polls the settled cache.