CB-WP-0016: the drop target that was never there

Provenance (tier S, one paragraph in lieu of survey and ADR): the human
check that kept INTENT stage 1 open was run and the drag was broken.
Root cause, worth more than the instance: drop targets were ids, and an
id must be unique, so exactly one element could ever be seat-0. The
relationship-graph circle took it and the seat card that every action
card's own text points at -- 'drag Attack onto a seat' -- silently had
none. A seat is drawn twice and both drawings are the seat; the document
model could not express that.

Drop keys are now data-drop. Any number of elements may carry the same
key, so a seat is droppable on its card and on its graph node. Measured
on a live server: seat-0/1/2 each appear twice, id survives only on
cb-status which is the one element the script looks up, and
down=action-attack&up=seat-1 returns ok.

Second defect: a drop on nothing returned without posting and without
touching the status line, so a broken target was indistinguishable from
a working page. resolve already refuses rather than defaulting, which is
right; refusing SILENTLY is not. The page now reports the raw fact --
'took action-attack, let go over nothing droppable' -- which names
elements, not moves, so ADR-0007 control 5 holds.

And the honest part: the general check added here -- every offered
affordance names a key that exists, driven through Policy::choose over
four real bot games -- does NOT catch the reported defect. seat-0 did
exist, on the graph circle. It is kept because a wholly absent target is
a real class, and paired with a targeted regression test that does catch
it. Three mutations, each red for its stated reason, including the
reported defect reintroduced; only the targeted test fires on that one.

A cb-play assertion matched id="action-ground" as a substring while
describing itself as checking the page; rewritten through drop_keys.

make all exits 0. Stage 1 stays open: verified by tests, mutation and a
live server, not by a human dragging.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

crates/cb-render-html/src/lib.rs

@@ -331,5 +331,184 @@ mod coverage {
     }
 }
 
+/// CB-WP-0016 T03: every affordance the page offers must name an element
+/// the page actually contains.
+///
+/// **This is the check that closes the class the human check found.** On
+/// 2026-08-02 the maintainer ran `cb-play --serve 0` and could not drag an
+/// action onto a seat. Every test in the repo was green. The cause: the
+/// visible seat cards carried no `id`, so `seat-{n}` existed only on the
+/// 26 px circles inside the relationship graph, while every action card
+/// read *"drag Attack onto a seat…"*.
+///
+/// Nothing could catch it. `jsrun::gesture` feeds element ids straight
+/// into a synthetic `{target:{id}}` and never hit-tests, so it establishes
+/// *"the script posts the ids it was given"* — never *"there is an element
+/// there to give."* The 42-path coverage gate asserts each view field is
+/// present in the parsed document, which a `<div>` with no id satisfies.
+///
+/// So: drive a real game, and at every decision point require that both
+/// halves of every offered affordance appear as real `id` attributes.
+#[cfg(test)]
+mod affordances {
+    use std::cell::RefCell;
+    use std::rc::Rc;
+
+    use cb_game_runtime::{Project, ScenarioGame, Setup, Viewer};
+    use cb_kernel::PlayerId;
+    use games_ground::bot::{play, Choice, Policy, RandomPolicy};
+    use games_ground::{GroundCommand, GroundState};
+
+    use crate::{doc, input};
+
+    fn fresh(seed: u64) -> GroundState {
+        GroundState::setup(
+            &Setup {
+                players: 3,
+                preset: "standard-3p".into(),
+                patch: std::collections::BTreeMap::new(),
+            },
+            seed,
+        )
+        .expect("a standard 3p deal")
+    }
+
+    /// Renders the page at every real decision point and checks it, then
+    /// delegates the actual choice.
+    ///
+    /// Hooking `Policy` rather than re-driving the game by hand matters:
+    /// these are the *same* decision points `cb-play --serve` renders at,
+    /// with the same `legal` list. A hand-rolled walk would be a second
+    /// implementation of the loop, and could agree with itself while
+    /// disagreeing with the thing shipped.
+    struct CheckingPolicy {
+        inner: RandomPolicy,
+        checked: Rc<RefCell<usize>>,
+    }
+
+    impl Policy for CheckingPolicy {
+        fn name(&self) -> &'static str {
+            "affordance-checking"
+        }
+
+        fn choose(
+            &mut self,
+            state: &GroundState,
+            seat: PlayerId,
+            legal: &[GroundCommand],
+            may_pass: bool,
+        ) -> Choice {
+            let view = state.project(Viewer::Player(seat));
+            let html = doc::document(&view, legal, "/command?t=x", Some(seat), may_pass);
+            let present = doc::drop_keys(&html);
+
+            for cmd in legal {
+                let Some((from, to)) = input::affordance(cmd, seat) else {
+                    // A command with no affordance is offered through the
+                    // numbered-button path instead. That is a stated
+                    // shape, not a missing element.
+                    continue;
+                };
+                assert!(
+                    present.contains(&from),
+                    "the page offers {cmd:?} whose GRAB id {from:?} is not an \
+                     element in the document (seat {seat:?}, step {:?})",
+                    state.step
+                );
+                assert!(
+                    present.contains(&to),
+                    "the page offers {cmd:?} whose DROP id {to:?} is not an \
+                     element in the document (seat {seat:?}, step {:?}). \
+                     Present ids: {present:?}",
+                    state.step
+                );
+            }
+            *self.checked.borrow_mut() += 1;
+            self.inner.choose(state, seat, legal, may_pass)
+        }
+    }
+
+    /// **The check that closes the class the human check found.**
+    ///
+    /// On 2026-08-02 the maintainer ran `cb-play --serve 0` and could not
+    /// drag an action onto a seat. Every test in the repo was green. The
+    /// cause: the visible seat cards carried no `id`, so `seat-{n}` existed
+    /// only on the 26 px circles inside the relationship graph, while every
+    /// action card read *"drag Attack onto a seat…"*.
+    ///
+    /// Nothing could catch it. `jsrun::gesture` feeds element ids straight
+    /// into a synthetic `{target:{id}}` and never hit-tests, so it
+    /// establishes *"the script posts the ids it was given"* — never
+    /// *"there is an element there to give."* The coverage gate asserts
+    /// each view field appears in the parsed document, which a `<div>` with
+    /// no id satisfies perfectly.
+    ///
+    /// An affordance naming an element that does not exist is the
+    /// harness-does-nothing shape in the presentation layer, and until now
+    /// it had no detector at all.
+    #[test]
+    fn every_offered_affordance_names_an_element_that_exists() {
+        let checked = Rc::new(RefCell::new(0usize));
+        for seed in 0..4u64 {
+            let mut policies: Vec<Box<dyn Policy>> = (0..3)
+                .map(|i| {
+                    Box::new(CheckingPolicy {
+                        inner: RandomPolicy::new(seed * 10 + i),
+                        checked: checked.clone(),
+                    }) as Box<dyn Policy>
+                })
+                .collect();
+            play(fresh(seed), &mut policies).expect("a bot game completes");
+        }
+        // Positive control: a run that rendered nothing would assert
+        // nothing and read as a pass — the exact failure this test exists
+        // to catch, one level up.
+        let n = *checked.borrow();
+        assert!(n >= 50, "checked only {n} decision point(s)");
+    }
+
+    /// **The regression test for the defect actually reported**, and the
+    /// reason the check above is not sufficient on its own.
+    ///
+    /// `every_offered_affordance_names_an_element_that_exists` passes on
+    /// the broken tree. `seat-0` *did* exist — on the 26 px circle in the
+    /// relationship graph — so an existence check over the whole document
+    /// cannot see that the seat *card*, which is what the instruction text
+    /// points at, was not droppable.
+    ///
+    /// A seat is drawn twice and both drawings are the seat. This asserts
+    /// the card specifically, by requiring the drop key on the element
+    /// that also carries `data-viewer` — the card, and nothing else.
+    #[test]
+    fn every_seat_card_is_a_drop_target_not_only_the_graph_node() {
+        let view = crate::testfix::view(Some(PlayerId(0)));
+        let html = doc::document(&view, &[], "/command?t=x", Some(PlayerId(0)), false);
+
+        for seat in view.players.keys() {
+            let card = format!(
+                "data-viewer=\"{}\" data-drop=\"seat-{}\"",
+                view.viewer == Some(*seat),
+                seat.0
+            );
+            assert!(
+                html.contains(&card),
+                "seat {seat:?} has a card that is not a drop target; looked for {card:?}"
+            );
+        }
+
+        // And the graph node keeps working — someone will have learned to
+        // aim at the circle, and this fix must not take that away.
+        let keys = doc::drop_keys(&html);
+        for seat in view.players.keys() {
+            assert!(keys.contains(&format!("seat-{}", seat.0)));
+        }
+        assert_eq!(
+            html.matches("data-drop=\"seat-0\"").count(),
+            2,
+            "seat 0 should be droppable in exactly two places: card and graph node"
+        );
+    }
+}
+
 #[cfg(test)]
 mod testfix;