diff --git a/LANGUAGE.md b/LANGUAGE.md
index 8509d175..23244163 100644
--- a/LANGUAGE.md
+++ b/LANGUAGE.md
@@ -1360,6 +1360,23 @@ checked. Diagnostics are limited to parse-level errors, reported as
 `file:line: error: message`; richer static checks (unknown identifiers,
 duplicate types, unknown fields, arity) are future work.
 
+`ludic-fmt --check` also enforces a project's style, stated in its `package.ludic`:
+
+```
+lint one_statement              # two statements on one line
+lint max_file_lines 100
+lint max_function_lines 50
+lint max_comment_lines 2        # a comment says why, in a line or two
+lint max_header_lines 3         # the comment that opens a file
+lint paths "src" "lab"          # what `ludic-fmt --lint` walks
+lint baseline "tests/lint-baseline.txt"
+```
+
+`ludic-fmt --lint` checks the project's paths; the baseline is a ratchet - the violations each file
+had when a rule came in, which it may keep but not add to, lowered automatically as they are
+fixed - so a rule can arrive in a codebase that breaks it today (`ludic-fmt --init-baseline`
+writes it). A `;` or a `#` inside a string does not count.
+
 ### Editors
 
 ```bash
diff --git a/changes/lint.md b/changes/lint.md
new file mode 100644
index 00000000..7a3e16ac
--- /dev/null
+++ b/changes/lint.md
@@ -0,0 +1,7 @@
+bump: minor
+type: feature
+**`ludic-fmt` enforces a project's style.** `lint` lines in `package.ludic` state it - one statement
+per line, file, function and comment-block limits (a file's opening comment its own) - and
+`ludic-fmt --check <files>` and `ludic-fmt --lint` (the project's paths) fail what breaks them, at the
+line. A baseline file is the ratchet that lets a rule arrive in an existing codebase: the counts a
+file had may stay but not grow, and they are lowered as they are fixed.
diff --git a/docs/language/app/app-monitor_count.md b/docs/language/app/app-monitor_count.md
new file mode 100644
index 00000000..0ee0bcc0
--- /dev/null
+++ b/docs/language/app/app-monitor_count.md
@@ -0,0 +1,20 @@
+---
+id: app-monitor_count
+name: App.monitor_count
+category: app
+kind: namespace-method
+tokens: App.monitor_count
+sig: App.monitor_count() -> int
+tip: How many displays the machine has.
+order: 49
+ns: App
+member: monitor_count
+---
+
+The number of displays attached, for a setting that chooses which one the game opens on; 1 on a platform that reports only the main one.
+
+```ludic
+program Screens {
+  entry { print(App.monitor_count()) }
+}
+```
diff --git a/docs/language/app/app-window_fixed.md b/docs/language/app/app-window_fixed.md
new file mode 100644
index 00000000..764dc429
--- /dev/null
+++ b/docs/language/app/app-window_fixed.md
@@ -0,0 +1,23 @@
+---
+id: app-window_fixed
+name: App.window_fixed
+category: app
+kind: namespace-method
+tokens: App.window_fixed
+sig: App.window_fixed(fixed)
+tip: A window that cannot be resized or maximised.
+order: 51
+ns: App
+member: window_fixed
+---
+
+With `true`, the window has no resize grip and no maximise button - for a launcher or a dialog, which a player should not take for the game itself.
+
+Parameters:
+- `fixed` — `true` to fix the window's size, `false` to free it
+
+```ludic
+program Launcher {
+  entry { App.window_fixed(true) }
+}
+```
diff --git a/docs/language/app/app-window_to_monitor.md b/docs/language/app/app-window_to_monitor.md
new file mode 100644
index 00000000..f9c8f1ab
--- /dev/null
+++ b/docs/language/app/app-window_to_monitor.md
@@ -0,0 +1,23 @@
+---
+id: app-window_to_monitor
+name: App.window_to_monitor
+category: app
+kind: namespace-method
+tokens: App.window_to_monitor
+sig: App.window_to_monitor(index)
+tip: Moves the window to that display.
+order: 50
+ns: App
+member: window_to_monitor
+---
+
+Moves the game's window onto display `index` (0 is the main one), keeping its size. A no-op where the platform leaves moving windows between screens to the player.
+
+Parameters:
+- `index` — which display, 0 to `App.monitor_count() - 1`
+
+```ludic
+program Move {
+  entry { App.window_to_monitor(0) }
+}
+```
diff --git a/docs/language/fs/fs-read_bytes.md b/docs/language/fs/fs-read_bytes.md
new file mode 100644
index 00000000..92704804
--- /dev/null
+++ b/docs/language/fs/fs-read_bytes.md
@@ -0,0 +1,26 @@
+---
+id: fs-read_bytes
+name: Fs.read_bytes
+category: fs
+kind: namespace-method
+tokens: Fs.read_bytes
+sig: Fs.read_bytes(path) -> []byte
+tip: A whole file as bytes, or null.
+order: 11
+ns: Fs
+member: read_bytes
+---
+
+Reads a whole file into a `[]byte` - bounds-checked, so no raw memory - or returns <code>null</code> when it cannot be read. It reads through the asset pack like every other load, so it finds a shipped game's files. Pair it with <code>text_of(b, n)</code> for text.
+
+Parameters:
+- `path` — the file to read
+
+```ludic
+program Header {
+  entry {
+    let b = Fs.read_bytes("sound.wav")
+    if b != null { print(b[24]) }
+  }
+}
+```
diff --git a/docs/language/fs/fs-write_bytes.md b/docs/language/fs/fs-write_bytes.md
new file mode 100644
index 00000000..c2b441cf
--- /dev/null
+++ b/docs/language/fs/fs-write_bytes.md
@@ -0,0 +1,30 @@
+---
+id: fs-write_bytes
+name: Fs.write_bytes
+category: fs
+kind: namespace-method
+tokens: Fs.write_bytes
+sig: Fs.write_bytes(path, data, count) -> bool
+tip: Writes the first count bytes of a []byte to a file.
+order: 12
+ns: Fs
+member: write_bytes
+---
+
+Writes the first `count` bytes of `data` to `path`, replacing the file, and answers whether all of them were written. `count` is clamped to the slice's length.
+
+Parameters:
+- `path` — the file to write
+- `data` — a `[]byte`, e.g. from <code>buffer(n)</code>
+- `count` — how many bytes of it
+
+```ludic
+program Save {
+  entry {
+    let b = buffer(2)
+    b[0] = 72
+    b[1] = 105
+    print(Fs.write_bytes("hi.bin", b, 2))
+  }
+}
+```
diff --git a/docs/language/input/input-text.md b/docs/language/input/input-text.md
new file mode 100644
index 00000000..36725770
--- /dev/null
+++ b/docs/language/input/input-text.md
@@ -0,0 +1,24 @@
+---
+id: input-text
+name: Input.text
+category: input
+kind: namespace-method
+tokens: Input.text
+sig: Input.text() -> string
+tip: The text typed this frame.
+order: 61
+ns: Input
+member: text
+---
+
+What the keyboard layout, the modifiers and any dead key produced this frame, as UTF-8 - the only right way to fill a name field, because it is the characters the player meant rather than the keys they pressed (a binding reads keys by position; text reads what they type). Empty in a headless run.
+
+```ludic
+program Name {
+  entry {
+    var name = ""
+    name = name + Input.text()
+    print(name)
+  }
+}
+```
diff --git a/docs/language/time/time-now_us.md b/docs/language/time/time-now_us.md
new file mode 100644
index 00000000..fb3cd4d5
--- /dev/null
+++ b/docs/language/time/time-now_us.md
@@ -0,0 +1,23 @@
+---
+id: time-now_us
+name: Time.now_us
+category: time
+kind: namespace-method
+tokens: Time.now_us
+sig: Time.now_us() -> long
+tip: A microsecond clock, for measuring.
+order: 7
+ns: Time
+member: now_us
+---
+
+Returns a steady microsecond count from the platform clock, for timing a piece of work - the difference between two readings is how long it took. It is not the game's time (that is <a href="time-now"><code>Time.now</code></a>) and it does not stop when the game is paused.
+
+```ludic
+program Timed {
+  entry {
+    let t0 = Time.now_us()
+    print(Time.now_us() - t0)
+  }
+}
+```
diff --git a/docs/language/time/time-sleep_us.md b/docs/language/time/time-sleep_us.md
new file mode 100644
index 00000000..2589326f
--- /dev/null
+++ b/docs/language/time/time-sleep_us.md
@@ -0,0 +1,26 @@
+---
+id: time-sleep_us
+name: Time.sleep_us
+category: time
+kind: namespace-method
+tokens: Time.sleep_us
+sig: Time.sleep_us(us)
+tip: Waits this many microseconds.
+order: 8
+ns: Time
+member: sleep_us
+---
+
+Suspends the program for about `us` microseconds - a high-resolution wait, for pacing a loop to a frame rate. No operating system's sleep is exact, so a limiter sleeps most of the wait and spins the last of it.
+
+Parameters:
+- `us` — how long to wait, in microseconds (a `long`)
+
+```ludic
+program Paced {
+  entry {
+    Time.sleep_us(1000)
+    print(1)
+  }
+}
+```
diff --git a/tools/ludic-cli/checks.ludic b/tools/ludic-cli/checks.ludic
index 5aeb8499..d5e95642 100644
--- a/tools/ludic-cli/checks.ludic
+++ b/tools/ludic-cli/checks.ludic
@@ -239,6 +239,30 @@ function collect_impl_pairs(allsrc: pointer, pairs: []pointer) -> void {
   }
 }
 
+# {ns}.{method} pairs a namespace declares with `alias` (L6) - the engine's in
+# runtime/native/namespaces.ludic - which the compiler dispatches as surely as its own
+function collect_alias_pairs(pairs: []pointer) -> void {
+  let src = read_file("runtime/native/namespaces.ludic")
+  if src == null { return }
+  var ns = ""
+  let n = slen(src)
+  var i = 0
+  while i < n {
+    let ln = s_trim(line_at(src, i))
+    i = i + slen(line_at(src, i)) + 1
+    if s_starts(ln, "namespace ") {
+      var e = 10
+      while e < slen(ln) and ln[e] != ' ' and ln[e] != '{' { e += 1 }
+      ns = sslice(ln, 10, e)
+    }
+    if s_starts(ln, "alias ") and slen(ns) > 0 {
+      var e = 6
+      while e < slen(ln) and ln[e] != '(' and ln[e] != ' ' and ln[e] != '=' { e += 1 }
+      set_add(pairs, ns + "." + sslice(ln, 6, e))
+    }
+  }
+}
+
 # gather the docs facts: token set, and the (ns.member) pairs of ns-method pages
 function collect_docs(tokens: []pointer, docpairs: []pointer) -> void {
   let list = capture("find docs/language -name '*.md' ! -name '_section.md' 2>/dev/null")
@@ -298,6 +322,7 @@ function cmd_check_impl() -> int {
   let allsrc = read_all_selfhost()
   let impl = new []pointer
   collect_impl_pairs(allsrc, impl)
+  collect_alias_pairs(impl)
   let tokens = new []pointer
   let docpairs = new []pointer
   collect_docs(tokens, docpairs)
diff --git a/tools/ludic-cli/tools.ludic b/tools/ludic-cli/tools.ludic
index 0d77c24f..1a041cb0 100644
--- a/tools/ludic-cli/tools.ludic
+++ b/tools/ludic-cli/tools.ludic
@@ -108,6 +108,26 @@ function cmd_test_tools() -> int {
   if shq(`{fmt} {tmp_dir()}/q.ludic | grep -q '\\[Pos, {{Foe}}\\]'`) { ok("query tags stay tight ({Foe})") } else { bad("query tags stay tight") }
   write_file(`{tmp_dir()}/al.ludic`, "program A {\n  const R_A:   int = 0    # one\n  const R_BBB: int = 1    # two\n}\n")
   if shq(`{fmt} {tmp_dir()}/al.ludic | grep -q 'R_A:   int = 0    # one'`) { ok("hand alignment preserved") } else { bad("hand alignment preserved") }
+  write_file(`{tmp_dir()}/gen.ludic`, "program G {\n  function f<T>(p: Pool<T>, m: Map<string, []int>) -> Pool<Pool<T>> {\n    let ok = a < b and c > d\n    return null\n  }\n}\n")
+  if shq(`{fmt} {tmp_dir()}/gen.ludic | cmp -s - {tmp_dir()}/gen.ludic`) { ok("type arguments stay together, comparisons keep their spaces (L5)") } else { bad("type arguments stay together (L5)") }
+
+  # --- the style rules a project states (L10) --------------------------------
+  let lp = `{tmp_dir()}/lintproj`
+  let lfmt = `{capture_line("pwd")}/{fmt}`       # the tests below run from inside the project
+  shell(`rm -rf {lp} && mkdir -p {lp}/src`)
+  write_file(`{lp}/package.ludic`, "package \"t\"\nlint one_statement\nlint max_file_lines 20\nlint max_function_lines 5\nlint max_comment_lines 2\nlint max_header_lines 3\nlint paths \"src\"\nlint baseline \"lint-baseline.txt\"\n")
+  write_file(`{lp}/src/a.ludic`, "# a.ludic - a file\nfunction f() -> int {\n  # one\n  # two\n  # three\n  let a = 1;  return a\n}\n")
+  if not shq(`cd {lp} && {lfmt} --lint > out.txt 2>&1`) and shq(`grep -q 'src/a.ludic:3: max_comment_lines' {lp}/out.txt`) and shq(`grep -q 'src/a.ludic:6: one_statement' {lp}/out.txt`) {
+    ok("--lint fails a comment over the limit and two statements on a line, at their lines")
+  } else { bad("--lint fails a comment over the limit and two statements on a line") }
+  if shq(`cd {lp} && {lfmt} --init-baseline > /dev/null && {lfmt} --lint > /dev/null`) { ok("--init-baseline allows what is there") } else { bad("--init-baseline allows what is there") }
+  write_file(`{lp}/src/b.ludic`, "function g() -> int {\n  let b = 2;  return b\n}\n")
+  if not shq(`cd {lp} && {lfmt} --lint > /dev/null 2>&1`) { ok("a new violation fails, whatever the baseline allows elsewhere") } else { bad("a new violation fails") }
+  write_file(`{lp}/src/b.ludic`, "function g() -> int { return 2 }\n")
+  write_file(`{lp}/src/a.ludic`, "# a.ludic - a file\nfunction f() -> int {\n  let a = 1;  return a\n}\n")
+  if shq(`cd {lp} && {lfmt} --lint > /dev/null && ! grep -q max_comment_lines lint-baseline.txt`) { ok("fixing one lowers the baseline, and it stays lowered") } else { bad("fixing one lowers the baseline") }
+  write_file(`{lp}/src/c.ludic`, "function h() -> string { return \"a; b\" }\n")
+  if shq(`cd {lp} && {lfmt} --lint > /dev/null`) { ok("a ; inside a string is not a statement") } else { bad("a ; inside a string is not a statement") }
 
   # --- --check exit status, for CI and pre-commit hooks ----------------------
   let chk1 = sh(`{fmt} --check -q {tmp_dir()}/m.ludic`)
diff --git a/tools/ludic-tools/fmt.ludic b/tools/ludic-tools/fmt.ludic
index 5d273075..ebb6e0d5 100644
--- a/tools/ludic-tools/fmt.ludic
+++ b/tools/ludic-tools/fmt.ludic
@@ -10,6 +10,9 @@
 #   ludic-fmt a.md               format the ```ludic fences in a document
 #   cat a.ludic | ludic-fmt -    filter mode (stdin -> stdout)
 program LudicFmt {
+  import "fmt_lint.ludic"
+  import "fmt_lint_rules.ludic"
+  import "fmt_lint_run.ludic"
   # ---- token kinds (mirror ludic_syntax.h) ----
   const LT_EOF: int = 0
   const LT_NL: int = 1
@@ -589,6 +592,8 @@ program LudicFmt {
     var quiet = false
     var changed = false
     var failed = false
+    var lint_only = false
+    var lint_init = false
     let files = new []pointer
     var ai = 1
     while ai < arg_count() {
@@ -597,11 +602,32 @@ program LudicFmt {
       else { if (a == "--check") or (a == "-l") { check = true }
       else { if (a == "-q") or (a == "--quiet") { quiet = true }
       else { if (a == "--indent") { ai += 1;  if ai < arg_count() { indent = 0;  let d = arg(ai);  var di = 0;  while d[di] != 0 { indent = indent * 10 + (d[di] - 48);  di += 1 } } }
-      else { if (a == "-h") or (a == "--help") { print("ludic-fmt — format Ludic source");  return }
-      else { push(files, a) } } } } }
+      else { if (a == "--lint") { lint_only = true }
+      else { if (a == "--init-baseline") { lint_only = true;  lint_init = true }
+      else { if (a == "-h") or (a == "--help") { print("ludic-fmt — format Ludic source; --check also checks package.ludic's lint rules, --lint checks the project");  return }
+      else { push(files, a) } } } } } } }
       ai += 1
     }
     if indent < 1 or indent > 8 { indent = 2 }
+    # --lint: the project's rules over its paths, against the baseline (L10)
+    if lint_only {
+      if not lint_config() { print("ludic-fmt --lint: package.ludic states no `lint` rules");  exit(2) }
+      let all = new []string
+      var li = 0
+      while li < len(lt_paths) {
+        lint_walk(lt_paths[li], all)
+        li += 1
+      }
+      var lj = 0
+      while lj < len(all) {
+        let t = read_file(all[lj])
+        if t != null { lint_file(all[lj], string(t)) }
+        lj += 1
+      }
+      if not lint_judge(lint_init) { exit(1) }
+      return
+    }
+    let linting = check and lint_config()
 
     # stdin filter mode
     if len(files) == 0 or (len(files) == 1 and (files[0] == "-")) {
@@ -620,6 +646,7 @@ program LudicFmt {
         file_write(file_stderr(), m, cstr_len(m));  failed = true
       } else {
         let is_md = ends_with(path, ".md") or ends_with(path, ".markdown")
+        if linting and not is_md { lint_file(path, string(text)) }
         let out = format_source(text, is_md, indent)
         let same = (out == text)
         if check {
@@ -638,6 +665,11 @@ program LudicFmt {
       fi += 1
     }
     if failed { exit(2) }
+    # the files named are part of a project: judge them, but never rewrite its baseline
+    if linting {
+      lt_partial = true
+      if not lint_judge(false) { exit(1) }
+    }
     if check and changed { exit(1) }
   }
 }
diff --git a/tools/ludic-tools/fmt_lint.ludic b/tools/ludic-tools/fmt_lint.ludic
new file mode 100644
index 00000000..7d7ffa5a
--- /dev/null
+++ b/tools/ludic-tools/fmt_lint.ludic
@@ -0,0 +1,111 @@
+# fmt_lint.ludic — L10: the style rules, checked by the formatter. A project states them in its
+# package.ludic, and `ludic-fmt --check` (the files given) or `ludic-fmt --lint` (the project's
+# paths) fails a file that breaks one:
+#   lint one_statement              two statements on one line (`a();  b()`)
+#   lint max_file_lines 100         a file's length
+#   lint max_function_lines 50      a function's, from `function` to its closing brace
+#   lint max_comment_lines 2        a block of comment lines - a comment says why, in a line or two
+#   lint max_header_lines 6         the block that opens a file, which says what the file is
+#   lint paths "src" "lab"          what --lint walks
+#   lint baseline "tests/lint-baseline.txt"
+# The baseline is a ratchet: `<count> <rule> <file>` lines, the violations a file had when a rule
+# came in. A file may have that many and no more; when it has fewer the baseline is lowered, and
+# `--init-baseline` writes it from what is there now.
+var lt_one: bool = false
+var lt_max_file: int = 0
+var lt_max_fn: int = 0
+var lt_max_comment: int = 0
+var lt_max_header: int = 0
+var lt_baseline: string = ""
+var lt_paths: []string = new []string
+var lr_file: []string = new []string
+var lr_rule: []string = new []string
+var lr_line: []int = new []int
+var lr_msg: []string = new []string
+
+function lt_atoi(s: string) -> int {
+  var n = 0
+  var i = 0
+  while i < len(s) {
+    let c = s[i]
+    if c >= '0' and c <= '9' { n = n * 10 + (c - '0') }
+    i += 1
+  }
+  return n
+}
+function lt_unquote(s: string) -> string {
+  let n = len(s)
+  if n >= 2 and s[0] == '"' and s[n - 1] == '"' { return s[1 .. n - 1] }
+  return s
+}
+# the words of a line, a "quoted" one kept whole, a # ending it
+function lt_words(line: string) -> []string {
+  let out = new []string
+  var i = 0
+  let n = len(line)
+  while i < n {
+    while i < n and (line[i] == ' ' or line[i] == 9) { i += 1 }
+    if i >= n or line[i] == '#' { break }
+    var j = i
+    if line[i] == '"' {
+      j = i + 1
+      while j < n and line[j] != '"' { j += 1 }
+      if j < n { j += 1 }
+    } else {
+      while j < n and line[j] != ' ' and line[j] != 9 { j += 1 }
+    }
+    push(out, lt_unquote(line[i .. j]))
+    i = j
+  }
+  return out
+}
+function lt_lines(text: string) -> []string {
+  let out = new []string
+  var a = 0
+  var i = 0
+  let n = len(text)
+  while i < n {
+    if text[i] == '\n' {
+      push(out, text[a .. i])
+      a = i + 1
+    }
+    i += 1
+  }
+  if a < n { push(out, text[a .. n]) }
+  return out
+}
+# the project's rules, from package.ludic in the directory ludic-fmt runs in
+function lint_config() -> bool {
+  let man = read_file("package.ludic")
+  if man == null { return false }
+  var any = false
+  let ls = lt_lines(string(man))
+  var i = 0
+  while i < len(ls) {
+    let w = lt_words(ls[i])
+    if len(w) >= 2 and w[0] == "lint" {
+      any = true
+      if w[1] == "one_statement" { lt_one = true }
+      if w[1] == "max_file_lines" and len(w) > 2 { lt_max_file = lt_atoi(w[2]) }
+      if w[1] == "max_function_lines" and len(w) > 2 { lt_max_fn = lt_atoi(w[2]) }
+      if w[1] == "max_comment_lines" and len(w) > 2 { lt_max_comment = lt_atoi(w[2]) }
+      if w[1] == "max_header_lines" and len(w) > 2 { lt_max_header = lt_atoi(w[2]) }
+      if w[1] == "baseline" and len(w) > 2 { lt_baseline = w[2] }
+      if w[1] == "paths" {
+        var k = 2
+        while k < len(w) {
+          push(lt_paths, w[k])
+          k += 1
+        }
+      }
+    }
+    i += 1
+  }
+  return any
+}
+function lt_report(path: string, rule: string, line: int, msg: string) -> void {
+  push(lr_file, path)
+  push(lr_rule, rule)
+  push(lr_line, line)
+  push(lr_msg, msg)
+}
diff --git a/tools/ludic-tools/fmt_lint_rules.ludic b/tools/ludic-tools/fmt_lint_rules.ludic
new file mode 100644
index 00000000..13c2898e
--- /dev/null
+++ b/tools/ludic-tools/fmt_lint_rules.ludic
@@ -0,0 +1,109 @@
+# fmt_lint_rules.ludic — L10: the rules themselves, over one file's text and its tokens
+function lint_file(path: string, text: string) -> void {
+  let ls = lt_lines(text)
+  if lt_max_file > 0 and len(ls) > lt_max_file { lt_report(path, "max_file_lines", len(ls), `{len(ls)} lines (the limit is {lt_max_file})`) }
+  lint_comments(path, ls)
+  lex(text)
+  if lt_one { lint_statements(path) }
+  if lt_max_fn > 0 { lint_functions(path) }
+}
+function lint_comments(path: string, ls: []string) -> void {
+  var run = 0
+  var start = 0
+  var code_seen = false
+  var i = 0
+  while i <= len(ls) {
+    var is_c = false
+    var blank = true
+    if i < len(ls) {
+      let w = lt_words_raw(ls[i])
+      is_c = w == 1
+      blank = w == 0
+    }
+    if is_c {
+      if run == 0 { start = i }
+      run += 1
+    } else {
+      if run > 0 {
+        let header = not code_seen and start == 0
+        if header and lt_max_header > 0 and run > lt_max_header { lt_report(path, "max_header_lines", start + 1, `the opening comment is {run} lines (the limit is {lt_max_header})`) }
+        if not header and lt_max_comment > 0 and run > lt_max_comment { lt_report(path, "max_comment_lines", start + 1, `a comment of {run} lines (the limit is {lt_max_comment}): say why in a line or two; the story goes in the commit or docs/`) }
+        run = 0
+      }
+      if not blank { code_seen = true }
+    }
+    i += 1
+  }
+}
+# 0 a blank line, 1 a comment line, 2 code
+function lt_words_raw(line: string) -> int {
+  var i = 0
+  while i < len(line) and (line[i] == ' ' or line[i] == 9) { i += 1 }
+  if i >= len(line) { return 0 }
+  if line[i] == '#' { return 1 }
+  return 2
+}
+function lint_statements(path: string) -> void {
+  var i = 0
+  while i < ntok() {
+    if tk_kind[i] == LT_OP and tok_len(i) == 1 and src[tk_start[i]] == ';' {
+      let k = tk_kind[i + 1]
+      if k != LT_NL and k != LT_COMMENT and k != LT_EOF { lt_report(path, "one_statement", tk_line[i] + 1, "two statements on one line") }
+    }
+    i += 1
+  }
+}
+function lint_functions(path: string) -> void {
+  var i = 0
+  while i < ntok() {
+    if tk_kind[i] == LT_KW and (tok_text(i) == "function") {
+      let end = lt_fn_end(i)
+      if end >= 0 {
+        let n = tk_line[end] - tk_line[i] + 1
+        if n > lt_max_fn { lt_report(path, "max_function_lines", tk_line[i] + 1, `{lt_fn_name(i)} is {n} lines (the limit is {lt_max_fn})`) }
+      }
+    }
+    i += 1
+  }
+}
+function lt_fn_name(i: int) -> string {
+  if i + 1 < ntok() { return string(tok_text(i + 1)) }
+  return "a function"
+}
+# the token closing a function's body, or -1 for one with none (an extern)
+function lt_fn_end(i: int) -> int {
+  var j = i + 1
+  var paren = 0
+  var closed = false
+  while j < ntok() {
+    let k = tk_kind[j]
+    if k == LT_OP and tok_len(j) == 1 {
+      let c = src[tk_start[j]]
+      if c == '(' { paren += 1 }
+      if c == ')' {
+        paren -= 1
+        if paren == 0 { closed = true }
+      }
+      if c == '{' and paren == 0 and closed { return lt_close(j) }
+    }
+    if (k == LT_NL or k == LT_EOF) and paren == 0 and closed { return -1 }
+    j += 1
+  }
+  return -1
+}
+function lt_close(open: int) -> int {
+  var depth = 0
+  var j = open
+  while j < ntok() {
+    if tk_kind[j] == LT_OP and tok_len(j) == 1 {
+      let c = src[tk_start[j]]
+      if c == '{' { depth += 1 }
+      if c == '}' {
+        depth -= 1
+        if depth == 0 { return j }
+      }
+    }
+    j += 1
+  }
+  return ntok() - 1
+}
diff --git a/tools/ludic-tools/fmt_lint_run.ludic b/tools/ludic-tools/fmt_lint_run.ludic
new file mode 100644
index 00000000..dced52b1
--- /dev/null
+++ b/tools/ludic-tools/fmt_lint_run.ludic
@@ -0,0 +1,117 @@
+# fmt_lint_run.ludic — L10: walking a project, and the baseline that lets a rule arrive in a
+# codebase that breaks it today without letting it get worse
+var bl_file: []string = new []string
+var bl_rule: []string = new []string
+var bl_count: []int = new []int
+var lt_partial: bool = false            # --check of some files: the baseline is read, never written
+
+function lint_walk(dir: string, out: []string) -> void {
+  let names = Fs.list(dir)
+  if names == null { return }
+  var i = 0
+  while i < len(names) {
+    let p = `{dir}/{names[i]}`
+    if Fs.is_dir(p) { lint_walk(p, out) }
+    else { if ends_with(p, ".ludic") { push(out, p) } }
+    i += 1
+  }
+}
+function lt_baseline_read() -> void {
+  if len(lt_baseline) == 0 { return }
+  let t = read_file(lt_baseline)
+  if t == null { return }
+  let ls = lt_lines(string(t))
+  var i = 0
+  while i < len(ls) {
+    let w = lt_words(ls[i])
+    if len(w) == 3 {
+      push(bl_count, lt_atoi(w[0]))
+      push(bl_rule, w[1])
+      push(bl_file, w[2])
+    }
+    i += 1
+  }
+}
+function lt_allowed(path: string, rule: string) -> int {
+  var i = 0
+  while i < len(bl_file) {
+    if bl_file[i] == path and bl_rule[i] == rule { return bl_count[i] }
+    i += 1
+  }
+  return 0
+}
+# how many violations of `rule` the file has now
+function lt_count(path: string, rule: string) -> int {
+  var n = 0
+  var i = 0
+  while i < len(lr_file) {
+    if lr_file[i] == path and lr_rule[i] == rule { n += 1 }
+    i += 1
+  }
+  return n
+}
+function lt_seen(fs: []string, rs: []string, path: string, rule: string) -> bool {
+  var i = 0
+  while i < len(fs) {
+    if fs[i] == path and rs[i] == rule { return true }
+    i += 1
+  }
+  return false
+}
+# true when every file is within its allowance; prints what is not
+function lint_judge(init: bool) -> bool {
+  lt_baseline_read()
+  let pf = new []string
+  let pr = new []string
+  var i = 0
+  while i < len(lr_file) {
+    if not lt_seen(pf, pr, lr_file[i], lr_rule[i]) {
+      push(pf, lr_file[i])
+      push(pr, lr_rule[i])
+    }
+    i += 1
+  }
+  var ok = true
+  var lowered = false
+  var total = 0
+  let out = buf_new()
+  buf_puts(out, "# lint baseline: the violations each file had when a rule came in (ludic-fmt --lint); only ever lowered\n")
+  var p = 0
+  while p < len(pf) {
+    let now = lt_count(pf[p], pr[p])
+    let allow = lt_allowed(pf[p], pr[p])
+    total += now
+    if not init and now > allow {
+      ok = false
+      lt_print_pair(pf[p], pr[p], now, allow)
+    }
+    var keep = now
+    if not init and now > allow { keep = allow }
+    if not init and now < allow { lowered = true }
+    if keep > 0 { buf_puts(out, `{keep} {pr[p]} {pf[p]}\n`) }
+    p += 1
+  }
+  var b = 0
+  while b < len(bl_file) {
+    if lt_count(bl_file[b], bl_rule[b]) == 0 and bl_count[b] > 0 and not init { lowered = true }
+    b += 1
+  }
+  if len(lt_baseline) > 0 and (init or lowered) and not lt_partial {
+    let f = file_open(lt_baseline, "wb")
+    if f != null {
+      let s = buf_str(out)
+      file_write(f, s, cstr_len(s))
+      file_close(f)
+    }
+  }
+  print(`  lint: {total} violation(s) in the baseline, {len(pf)} file/rule pair(s)`)
+  return ok
+}
+function lt_print_pair(path: string, rule: string, now: int, allow: int) -> void {
+  var i = 0
+  while i < len(lr_file) {
+    if lr_file[i] == path and lr_rule[i] == rule { print(`{path}:{lr_line[i]}: {rule}: {lr_msg[i]}`) }
+    i += 1
+  }
+  if allow > 0 { print(`  ({path} may have {allow} of {rule}, and has {now})`) }
+}