feat: drop the kettle plugin; the binary writes its own skills

The plugin and the binary shipped on two release cadences and nothing on an
operator's machine ever checked that the one they installed described the other.
The generated flag block existed precisely so a renamed flag could not ship with
documentation recommending the old one — and then shipped one version behind the
registry it came from, which is the same bug one hop downstream.

So the prose moved into the binary. `internal/scaffold` embeds every document;
`kettle init` and `kettle gen scaffold` write them into a project's own
`.claude/`. The two cannot disagree because there is one artefact.

The namespace survived the move. A project's skills are flat, so the prefix is
spelled into the directory name (`kettle-issue`); a project's *commands* take
their namespace from a subdirectory, so `commands/kettle/init.md` is still
`/kettle:init`. Four of the six command files are thin pointers at a skill, and
that is what kept ~1,600 lines of `/kettle:…` cross-references true without a
rewrite. `init` and `auth` lost `disable-model-invocation: true` — being a
command is that property — and `auth` now restricts `allowed-tools` so a model
cannot reach `kettle auth add` at all.

`gen scaffold` writes files whole rather than splicing a region. The old
refusal protected somebody's hand-written prose around the block; that prose is
embedded now, so there is none to protect, and preserving local edits would
freeze a project's documentation at whatever version first initialized it.
`--check` warns before an upgrade discards one.

The plugin's `agents-sync.sh` — 141 lines of Python behind a filename that said
`.sh` — became `internal/mirror` and `kettle mirror`. Same seven branches, same
refusal to merge two real files that differ, now with a table test per branch
and a check that a repair converges in one pass. `--hook` is the PreToolUse
form and exits 0 on every path including a panic. It is opt-in per project,
which is strictly narrower than the plugin hook that was on for everybody who
installed it.

`kettle init --interactive` walks a person through the login, the token (read
with the echo off, so it lands in no history and no file), the repository, the
`.claude/` tree and the mirror hook. It refuses a stdin that is not a terminal
and names the flags instead: every question it asks has one, and it performs
nothing itself, so an interactive run and a flag run are one code path.

Two rules that used to be prose are now the binary's: init refuses a linked
worktree and names the main checkout, and writing into an existing
`.claude/settings.json` is refused with the snippet printed rather than
reformatting a file the operator commits.

The scaffold version stamp went to its own `.kettle/scaffold.yaml` rather than
into `config.yaml`, because unknown keys there are a hard error and that file
may be committed and read by whatever build each machine has.

golang.org/x/term becomes a direct dependency; it was already in the tree
indirectly, so no module was added.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

cli/internal/scaffold/scaffold.go

@@ -0,0 +1,131 @@
+// Package scaffold holds the documents kettle writes into a project: the slash
+// commands an operator invokes, the skills a model loads, and the runner
+// subagent.
+//
+// They live inside the binary rather than beside it. A plugin shipped this prose
+// once, on its own release cadence, and nothing on an operator's machine checked
+// that the plugin they had installed described the binary they had installed —
+// so a renamed flag could ship with documentation recommending the old one,
+// which is the exact failure the generated block was invented to prevent, one
+// hop further downstream. Prose that travels inside the binary cannot be a
+// version behind it.
+//
+// This package depends on nothing but the standard library and holds no
+// rendering logic: it hands out embedded files and says which of them carry a
+// generated region. Splicing the command registry into that region is
+// internal/cmd's, because the registry is.
+package scaffold
+
+import (
+	"embed"
+	"io/fs"
+	"path"
+	"sort"
+	"strings"
+)
+
+//go:embed all:assets
+var assets embed.FS
+
+const assetRoot = "assets"
+
+// Marker is the directory these files are written into, relative to the project
+// root. It belongs to the agent harness, not to kettle: everything kettle owns
+// is under `.kettle/`, and this is the one tree it writes that somebody else
+// defines the shape of.
+const Marker = ".claude"
+
+// generated maps an output path to the command group whose flag table belongs
+// in it. A file that is not in here carries no generated region and is shipped
+// exactly as embedded.
+//
+// Explicit rather than derived from the directory name: the group ⇄ skill
+// correspondence is a decision, and one that has not always held — `init` and
+// `auth` are commands with no skill of their own, and `project` is a skill
+// covering four commands. A test in internal/cmd asserts every group in the
+// registry is named here exactly once, so adding a group fails loudly rather
+// than silently shipping a skill nobody can find.
+var generated = map[string]string{
+	"skills/kettle-project/SKILL.md": "project",
+	"skills/kettle-issue/SKILL.md":   "issue",
+	"skills/kettle-sync/SKILL.md":    "sync",
+	"skills/kettle-api/SKILL.md":     "api",
+}
+
+// File is one document, ready to be written under the output directory.
+type File struct {
+	// Path is relative to the output directory, always with forward slashes:
+	// "commands/kettle/init.md", "skills/kettle-issue/SKILL.md".
+	Path string
+	// Body is the file as embedded — before any generated region is spliced in.
+	Body string
+	// Group is the command group whose flag table belongs in this file, or "".
+	Group string
+}
+
+// Files is every document, sorted by path.
+//
+// Sorted, not in walk order, because the sort is the promise: two runs of the
+// same binary produce the same list, so a receipt and a --check diff are
+// comparable between machines.
+func Files() []File {
+	var out []File
+	_ = fs.WalkDir(assets, assetRoot, func(p string, d fs.DirEntry, err error) error {
+		if err != nil || d.IsDir() {
+			return err
+		}
+		rel := strings.TrimPrefix(p, assetRoot+"/")
+		body, err := assets.ReadFile(p)
+		if err != nil {
+			return err
+		}
+		out = append(out, File{Path: rel, Body: string(body), Group: generated[rel]})
+		return nil
+	})
+	sort.Slice(out, func(i, j int) bool { return out[i].Path < out[j].Path })
+	return out
+}
+
+// Groups lists every command group that has a file here, sorted.
+func Groups() []string {
+	out := make([]string, 0, len(generated))
+	for _, g := range generated {
+		out = append(out, g)
+	}
+	sort.Strings(out)
+	return out
+}
+
+// PathFor is the output path carrying a group's flag table, or "".
+func PathFor(group string) string {
+	for p, g := range generated {
+		if g == group {
+			return p
+		}
+	}
+	return ""
+}
+
+// Dirs lists the directories the output tree is made of, parents first, so a
+// caller can create them in order.
+func Dirs() []string {
+	seen := map[string]bool{}
+	var out []string
+	for _, f := range Files() {
+		for _, d := range parents(path.Dir(f.Path)) {
+			if !seen[d] {
+				seen[d] = true
+				out = append(out, d)
+			}
+		}
+	}
+	sort.Strings(out)
+	return out
+}
+
+func parents(dir string) []string {
+	if dir == "." || dir == "" {
+		return nil
+	}
+	return append(parents(path.Dir(dir)), dir)
+}