diff --git a/src/detect.jl b/src/detect.jl
index 059f7d4..5f8a1fc 100644
--- a/src/detect.jl
+++ b/src/detect.jl
@@ -207,6 +207,20 @@ _probe(flag) = :($flag[] || ($flag[] = true))
 # marked definition. `record`'s call paths are built out of exactly that.
 function _instrument(def, flag, src::LineNumberNode)
     def isa Expr || return nothing
+    if def.head === :macrocall
+        # An annotating macro — `@inline` and its neighbours — leaves the body alone, so the probe
+        # rides inside the definition it wraps and the wrapper is rebuilt around the result. Only
+        # those reach here: `_subject` marks a macrocall instrumentable exactly when the macro is
+        # in `_ANNOTATING_MACROS`, and refuses or opts out of every other one.
+        #
+        # Returning `nothing` here instead — which is what this did — did not merely lose the
+        # observation. The flag is registered either way, so `@experimental "…" @inline f(x) = x`
+        # counted as an observable definition that no call could ever set: `entered` reported it
+        # as not entered no matter what ran, and `unverified` reported it forever.
+        inner = _instrument(def.args[end], flag, src)
+        inner === nothing && return nothing
+        return Expr(:macrocall, def.args[1:(end - 1)]..., inner)
+    end
     (def.head === :function || def.head === :(=)) || return nothing
     _is_signature(def.args[1]) || return nothing
     length(def.args) == 2 || return nothing
diff --git a/src/mark.jl b/src/mark.jl
index 94ce547..de65b76 100644
--- a/src/mark.jl
+++ b/src/mark.jl
@@ -259,9 +259,10 @@ The difference shows up in three places: [`reach`](@ref) only reports the marked
 # What gets observed
 
 A definition **with a body** — `function` and short-form `f(x) = …`, including parametric,
-callable-object and return-type-annotated signatures — also gets a flag in that body, so
-[`entered`](@ref) can report whether the run went through it. The flag is one short-circuit read:
-`1.03x` on one thread and `0.985x` on eight, measured over 10M calls of a numeric body.
+callable-object and return-type-annotated signatures, and those wrapped in an annotating macro
+such as `@inline` — also gets a flag in that body, so [`entered`](@ref) can report whether the run
+went through it. The flag is one short-circuit read: `1.03x` on one thread and `0.985x` on eight,
+measured over 10M calls of a numeric body.
 
 Every other form is a **declaration only** — recorded, queryable and audited, but not observed:
 
@@ -272,8 +273,9 @@ Every other form is a **declaration only** — recorded, queryable and audited,
 | a name list (`@experimental "…" a b c`) | no — those definitions are elsewhere and already compiled |
 | `struct`, `abstract type`, `primitive type`, `const`, assignment | no — nothing is *entered* |
 | `macro` | no |
+| `@inline`, `@noinline`, `@propagate_inbounds`, `@assume_effects`, `@constprop`, `@nospecializeinfer` | yes — they annotate a definition and leave its body alone, so the flag rides inside |
 | `@generated function` | no — the body returns an expression, so a probe there would be generated rather than run |
-| `Base.@kwdef`, `@inline`, `@noinline` and the other pass-through macros | no |
+| `Base.@kwdef` | no — it wraps a `struct`, which has no body |
 
 One flag per marked **name**, not per method: flags are name-keyed, so two methods of a marked
 name share one.
diff --git a/test/spec/README.md b/test/spec/README.md
index ef39068..dd060db 100644
--- a/test/spec/README.md
+++ b/test/spec/README.md
@@ -49,13 +49,13 @@ that is entirely `@test_broken` is a claim written down, not a check being run.
 | `test_spec_dispatch.jl` | 15 | 15 | 0 | one call site, several methods, only some marked — the branch |
 | `test_spec_docstring.jl` | 9 | 9 | 0 | a mark and a docstring are different accounts and must coexist |
 | `test_spec_foreign.jl` | 13 | 13 | 0 | marking a method on somebody else's generic — the `QAtlas.fetch` case |
-| `test_spec_forms.jl` | 24 | 24 | 0 | the definition forms a real package hits on its second afternoon |
+| `test_spec_forms.jl` | 28 | 28 | 0 | the definition forms a real package hits on its second afternoon |
 | `test_spec_integration.jl` | 19 | 19 | 0 | where the mark has to surface: docs, Aqua, releases, provenance, CI |
 | `test_spec_lifecycle.jl` | 16 | 16 | 0 | the mark's EXIT, and an entry point that is a module rather than a function |
 | `test_spec_profile.jl` | 45 | 45 | 0 | what a real run went through, how often, and how much of it |
 | `test_spec_propagate.jl` | 20 | 20 | 0 | a caller that never names a marked thing still depends on it |
 | `test_spec_verify.jl` | 9 | 9 | 0 | how well is a marked thing exercised by the tests |
-| **10 files** | **182** | **182** | **0** | |
+| **10 files** | **186** | **186** | **0** | |
 <!-- END GENERATED -->
 
 The table is generated and pinned by `test/test_spec_table.jl`, which fails if it goes stale —
diff --git a/test/spec/test_spec_forms.jl b/test/spec/test_spec_forms.jl
index 8fc408f..4dad829 100644
--- a/test/spec/test_spec_forms.jl
+++ b/test/spec/test_spec_forms.jl
@@ -159,14 +159,51 @@ end
 end
 
 @testset "@inline and the mark compose in both orders" begin
-    @test begin
-        @eval module InlineMarked
-        using ExperimentalAPI
-        @experimental "kernel unverified" @inline f(x) = x
-        @inline @experimental "kernel unverified" g(x) = x
-        end
-        Set([mk.name for mk in experimental(Main.InlineMarked)]) == Set([:f, :g])
+    # "Compose" was asserted as `Set([:f, :g])` — the names are marked — and that is satisfied by
+    # a mark that can never fire. Measured: the two orders did NOT compose the same way. With the
+    # mark outside, `_instrument` saw a `:macrocall` and returned `nothing`, so the flag was
+    # registered and nothing ever set it; with the mark inside, the flag worked. `entered` said
+    # `[:g]` after calling both. An `@inline` kernel is exactly what this package is for, so the
+    # claim has to be about the observation, not about the name.
+    @eval module InlineMarked
+    using ExperimentalAPI
+    @experimental "kernel unverified" @inline f(x) = x
+    @inline @experimental "kernel unverified" g(x) = x
+    end
+    @test Set([mk.name for mk in experimental(Main.InlineMarked)]) == Set([:f, :g])
+    @test Set(p.name for p in ExperimentalAPI.probes(Main.InlineMarked)) == Set([:f, :g])
+    Main.InlineMarked.f(1)
+    Main.InlineMarked.g(1)
+    @test Set(e.name for e in ExperimentalAPI.entered(Main.InlineMarked)) == Set([:f, :g])
+end
+
+@testset "every annotating macro carries the flag, and the opaque ones still do not" begin
+    # The control the testset above cannot be on its own: a fix that instrumented EVERY macrocall
+    # would satisfy it while putting a probe inside `@generated`'s returned expression, where it
+    # is generated rather than run. The split between the two lists is the claim.
+    @eval module Annotated
+    using ExperimentalAPI
+    @experimental "a" @inline a(x) = x
+    @experimental "b" @noinline b(x) = x
+    @experimental "c" Base.@propagate_inbounds c(x) = x
+    @experimental "d" Base.@assume_effects :terminates_locally d(x) = x
+    @experimental "e" @generated e(x) = :(x)
+    @experimental "f" Base.@kwdef struct Opaque
+        n::Int = 1
+    end
     end
+    for n in (:a, :b, :c, :d)
+        Core.eval(Main.Annotated, :($n(1)))
+    end
+    Main.Annotated.e(1)
+    Main.Annotated.Opaque()
+    @test Set(e.name for e in ExperimentalAPI.entered(Main.Annotated)) ==
+        Set([:a, :b, :c, :d])
+    # …and the two that are not observed are still marked, queryable and audited.
+    @test Set(mk.name for mk in experimental(Main.Annotated)) ==
+        Set([:a, :b, :c, :d, :e, :Opaque])
+    @test Set(p.name for p in ExperimentalAPI.probes(Main.Annotated)) ==
+        Set([:a, :b, :c, :d])
 end
 
 @testset "a definition produced by @eval can be marked by name" begin
@@ -259,6 +296,86 @@ end
     @test !occursin("mark.jl", msg)
 end
 
+# Each of these three is refused today, and none of the messages was pinned by anything — measured
+# 2026-09-08 by grepping `test/` for their text and finding zero hits. A refusal is half of what
+# this macro does: the other half of "cannot guess which name this defines" is saying what to
+# write instead, and a message can rot into a bare failure without a single test going red.
+@testset "a wrapping macro this cannot read is refused by name, and points at the name list" begin
+    for (label, body) in (
+        "another @experimental" => """@experimental "outer" @experimental "inner" f(x) = x""",
+        "an unknown macro" => """@experimental "why" @assert true""",
+    )
+        @testset "$label" begin
+            e = try
+                include_string(
+                    Main,
+                    "module Wrap_$(hash(label))\nusing ExperimentalAPI\n$body\nend",
+                    "wrap.jl",
+                )
+                nothing
+            catch err
+                err
+            end
+            # Unwrapped in a loop, not once: a macro that throws while expanding a `module` body
+            # passed through `include_string` comes back wrapped TWICE, and a single `.error`
+            # leaves a `LoadError` that reads exactly like the failure it is hiding.
+            while e isa LoadError
+                e = e.error
+            end
+            @test e isa ArgumentError
+            msg = sprint(showerror, e)
+            # Names the macro it could not read, so the author knows which line to change…
+            @test occursin("@", msg)
+            # …and the form that always works, which is what makes it actionable.
+            @test occursin("@experimental \"why\" the_name", msg)
+        end
+    end
+end
+
+@testset "a block of definitions is refused rather than half-marked" begin
+    # The dangerous silence: `begin f(x)=x; g(x)=x end` has two names and the macro can only
+    # record one. Marking the first and dropping the second would be a covenant that omits a
+    # definition without saying so.
+    e = try
+        @eval module BlockSubject
+        using ExperimentalAPI
+        @experimental "why" begin
+            f(x) = x
+            g(x) = x
+        end
+        end
+        nothing
+    catch err
+        err isa LoadError ? err.error : err
+    end
+    @test e isa ArgumentError
+    msg = sprint(showerror, e)
+    @test occursin("block", msg)
+    @test occursin("the_name", msg)
+end
+
+@testset "a bare qualified name is refused, because the module it names is not ours to mark" begin
+    # `@experimental "why" Sub.g` reads as marking somebody else's `g`. The name list records into
+    # the module the macro ran in, so accepting it would file the mark in the wrong registry.
+    e = try
+        @eval module QualifiedBare
+        using ExperimentalAPI
+        module Sub
+            g(x) = x
+        end
+        @experimental "why" Sub.g
+        end
+        nothing
+    catch err
+        err isa LoadError ? err.error : err
+    end
+    @test e isa ArgumentError
+    msg = sprint(showerror, e)
+    @test occursin("Sub.g", msg)        # the expression the author wrote, not a generic complaint
+    @test occursin("WHICH method", msg)  # why a bare qualified name is not enough…
+    @test occursin("Sub.g(::", msg)      # …and the form that is, spelled out with their own name
+end
+
 # ── metadata ─────────────────────────────────────────────────────────────────────────────────
 
 @testset "since must be a version, and the refusal must say so" begin