# Hyperscript conformance → 100%

Goal: take the hyperscript upstream conformance suite from **1213/1496 (81%)** to a clean 100%. Queue-driven — single-agent loop on `architecture` branch, one cluster per commit.

## North star

```
Baseline: 1213/1496 (81.1%)
Target:   1496/1496
Gap:      283 tests  (130 real fails + 153 SKIPs)
```

Track after each iteration via `mcp__hs-test__hs_test_run` on the relevant suite, not the whole thing (full runs take 10+min and include hanging tests — 196/199/200/615/1197/1198 hang under the 200k step limit).

## How to run tests

```
mcp__hs-test__hs_test_run(suite="hs-upstream-<cluster>")        # fastest, one suite
mcp__hs-test__hs_test_run(start=0, end=195)                     # early range
mcp__hs-test__hs_test_run(start=201, end=614)                   # mid range (skip hypertrace hangs)
mcp__hs-test__hs_test_run(start=616, end=1196)                  # late-1, skip repeat-forever hangs
mcp__hs-test__hs_test_run(start=1199)                           # late-2 after hangs
```

## File layout

Runtime/compiler/parser live in `lib/hyperscript/*.sx`. The test runner at `tests/hs-run-filtered.js` loads `shared/static/wasm/sx/hs-*.sx` — **after every `.sx` edit you must `cp lib/hyperscript/<file>.sx shared/static/wasm/sx/hs-<file>.sx`**.

The test fixtures live in `spec/tests/test-hyperscript-behavioral.sx`, generated from `tests/playwright/generate-sx-tests.py`. **Never edit the behavioral.sx fixture directly** — fix the generator or the runtime.

## Cluster queue

Each cluster below is one commit. Order is rough — a loop agent may skip ahead if a predecessor is blocked. **Status:** `pending` / `in-progress` / `done (+N)` / `blocked (<reason>)`.

### Parallel-worktree mode

When fanning out multiple clusters at once (Agent with `isolation: "worktree"`), each worktree agent:

1. Works on a fresh copy of the repo — no contention on the mutable tree.
2. Picks **one** cluster, runs the full loop (read, baseline, fix, sync WASM, verify smoke 0-195 + target suite, commit on the worktree's branch).
3. Leaves its branch + commit SHA for the orchestrator. **Does not push**, does not update `plans/hs-conformance-to-100.md` or the scoreboard — those updates happen in the orchestrator's cherry-pick commit so the ledger stays linear.
4. Scope inside the worktree is unchanged (`lib/hyperscript/**`, `shared/static/wasm/sx/hs-*`, `tests/hs-run-filtered.js`, `tests/playwright/generate-sx-tests.py` + its regen output `spec/tests/test-hyperscript-behavioral.sx`). Do **not** edit the plan or the scoreboard inside the worktree — that's the orchestrator's job.

Orchestrator cherry-picks worktree commits onto `architecture` one at a time; resolves conflicts as they arrive (most will be trivial since each cluster lives in its own parser/compiler branch or in a different mock).

**Good candidates to parallelise:** clusters that touch disjoint surfaces — e.g. 26 (resize observer) and 27 (intersection observer) edit the same mock file but different class stubs; 25 (parenthesised commands) is parser-only; 30 (logAll config) is bootstrap/integration-only. Avoid fanning out clusters that all rewrite the same dispatch spot (`emit-set`, `parse-expr`) in the same commit.

**Cherry-pick footgun (observed 2026-04-24):** `sx-tree`'s pretty-printer reformats large regions when an edit lands in the middle of a big `let`/`fn` body. Two worktree commits whose logical diffs touch *different* defines in the same `.sx` file will still conflict textually because the pretty-print shuffles comments and indentation. Because `.sx` files can't be `Edit`-ed (hook blocks `Edit`/`Write`), conflict markers left by git are unrepairable. **Workaround:** when you see a conflict, abort the cherry-pick and re-apply the worktree commit surgically via `sx_replace_node`/`sx_insert_near` on the specific paths that changed. The logical diff is usually small (5–10 nodes); read it with `git show SHA file.sx` and apply it as a series of tree edits on top of current HEAD.

### Bucket A: runtime fixes, single-file (low risk, high yield)

1. **[done (+4)] fetch JSON unwrap** — `hs-upstream-fetch` 4 tests (`can do a simple fetch w/ json` + 3 variants) got `{:__host_handle N}`. Root: `hs-fetch` in `runtime.sx` returns raw host Response object instead of parsing JSON body. Fix: when format is `"json"`, unwrap via `host-get "_json"` and `json-parse`. Expected: +4.

2. **[done (+1)] element → HTML via outerHTML** — `asExpression / converts an element into HTML` (1 test) + unlocks response fetches. Mock DOM `El` class in `tests/hs-run-filtered.js` has no `outerHTML` getter. Add a getter computed from `tagName` + `attributes` + `children` (recurse). Expected: +1 direct, + knock-on in fetch.

3. **[done (+2)] Values dict insertion order** — `asExpression / Values | FormEncoded` + `| JSONString` (2 tests) — form fields come out `lastName, phone, firstName, areaCode`. Root: `hs-values-absorb` in `runtime.sx` uses `dict-set!` but keys iterate in non-insertion order. Investigate `hs-gather-form-nodes` walk — the recursive `kids` traversal silently fails when `children` is a JS Array (not sx-list), so nested inputs arrive via a different path. Fix: either coerce children to sx-list at the gather boundary OR rewrite gather to explicitly use sx-level iteration helpers. Expected: +2.

4. **[done (+3)] `not` precedence over `or`** — `expressions/not` 3 tests (`not has higher precedence than or`, `not with numeric truthy/falsy`, `not with string truthy/falsy`). Check parser precedence — `not` should bind tighter than `or`. Fix in `parser.sx` expression-level precedence. Expected: +3.

5. **[done (+1)] `some` selector for nonempty match** — `expressions/some / some returns true for nonempty selector` (1 test). `some .class` probably returns the list, not a boolean. Runtime fix. Expected: +1.

6. **[done (+2)] string template `${x}`** — `expressions/strings / string templates work w/ props` + `w/ braces` (2 tests). Template interpolation isn't substituting property accesses. Check `hs-template` runtime. Expected: +2.

7. **[done (+1) — partial, 3 tests remain: inserted-button handler doesn't fire for afterbegin/innerHTML paths; might need targeted trace of hs-boot-subtree! or _setInnerHTML timing] `put` hyperscript reprocessing** — `put / properly processes hyperscript at end/start/content/symbol` (4 tests, all `Expected 42, got 40`). After a put operation, newly inserted HS scripts aren't being activated. Fix: `hs-put-at!` should `hs-boot-subtree!` on the target after DOM insertion. Expected: +4.

8. **[done (+1)] `select returns selected text`** (1 test, `hs-upstream-select`). Runtime `hs-get-selection` helper reads `window.__test_selection` stash (or falls back to real `window.getSelection().toString()`). Compiler rewrites `(ref "selection")` to `(hs-get-selection)`. Generator detects the `createRange` / `setStart` / `setEnd` / `addRange` block and emits a single `(host-set! ... __test_selection ...)` op with the resolved text slice of the target element. Expected: +1.

9. **[done (+4)] `wait on event` basics** — `wait / can wait on event`, `on another element`, `waiting ... sets it to the event`, `destructure properties in a wait` (4 tests). Event-waiter suspension issue. Expected: +3-4.

10. **[done (+1)] `swap` variable ↔ property** — `swap / can swap a variable with a property` (1 test). Swap command doesn't handle mixed var/prop targets. Expected: +1.

11. **[done (+4)] `hide` strategy** — `hide / can configure hidden as default`, `can hide with custom strategy`, `can set default to custom strategy`, `hide element then show element retains original display` (4 tests). Strategy config plumbing. Expected: +3-4.

12. **[done (+2)] `show` multi-element + display retention** — `show / can show multiple elements with inline-block`, `can filter over a set of elements using the its symbol` (2 tests). Expected: +2.

13. **[done (+2) — partial, `can toggle for a fixed amount of time` needs an async mock scheduler (sync io-sleep collapses the toggle/un-toggle into one click frame)] `toggle` multi-class + timed + until-event** — `toggle` (3 assertion-fail tests). Expected: +3.

14. **[done (+1)] `unless` modifier** — `unlessModifier / unless can conditionally execute` (1 test). Parser/compiler addition. Expected: +1.

15. **[done (+2) — partial, `can use initial to transition to original value` needs `on click N` count-filtered events (same sync-mock block as clusters 11/13)] `transition` query-ref + multi-prop + initial** — `transition` 3 tests. Expected: +2-3.

16. **[done (+1)] `send can reference sender`** — 1 assertion fail. Expected: +1.

17. **[blocked: tell semantics are subtle — `me` should stay as the original element for explicit `to me` writes but the implicit default for bare `add .bar` inside `tell X` should be X. Attempted just leaving `you`/`yourself` scoped (dropping the `me` shadow) regressed 4 passing tests (`restores proper implicit me`, `works with an array`, etc.) which rely on bare commands using `me` as told-target. Proper fix requires a `beingTold` symbol distinct from `me`, with bare commands compiling to `beingTold-or-me` and explicit `me` always the original — more than a 30-min cluster budget.] `tell` semantics** — `tell / attributes refer to the thing being told`, `does not overwrite me symbol`, `your symbol represents thing being told` (3 tests). Expected: +3.

18. **[done (+2)] `throw respond async/sync`** — `throw / can respond to async/sync exceptions in event handler` (2 tests). Expected: +2.

### Bucket B: parser/compiler additions (medium risk, shared files)

19. **[done (+13)] `pick` regex + indices** — `pick` 13 tests. Regex match, flags, `of` syntax, start/end, negative indices. Big enough that a single commit might fail — break into pick-regex and pick-indices if needed. Expected: +10-13.

20. **[done (+3)] `repeat` property for-loops + where** — `repeat / basic property for loop`, `can nest loops`, `where clause can use the for loop variable name` (3 tests). Expected: +3.

21. **[done (+1)] `possessiveExpression` property access via its** — `possessive / can access its properties` (1 test, Expected `foo` got ``). Expected: +1.

22. **[done (+1)] window global fn fallback** — `regressions / can invoke functions w/ numbers in name` + `can refer to function in init blocks`. Added `host-call-fn` FFI primitive (commit 337c8265), `hs-win-call` runtime helper, simplified compiler emit (direct hs-win-call, no guard), `def` now also registers fn on `window[name]`. Generator: fixed `\"` escaping in hs-compile string literals. Expected: +2-4.

23. **[done (+1)] `me symbol works in from expressions`** — `regressions` (1 test, Expected `Foo`). Check `from` expression compilation. Expected: +1.

24. **[done (+1)] `properly interpolates values 2`** — URL interpolation regression (1 test). Likely template string + property access. Expected: +1.

25. **[done (+1)] `can support parenthesized commands and features`** — `parser` (1 test, Expected `clicked`). Parser needs to accept `(cmd...)` grouping in more contexts. Expected: +1.

### Bucket C: feature stubs (DOM observer mocks)

26. **[done (+3)] resize observer mock + `on resize`** — 3 tests. Add a minimal `ResizeObserver` mock to `hs-run-filtered.js`, plus parse/compile `on resize`. Expected: +3.

27. **[done (+3)] intersection observer mock + `on intersection`** — 3 tests. Mock `IntersectionObserver`; compile `on intersection` with margin/threshold modifiers. Expected: +3.

28. **[done (+4)] `ask`/`answer` + prompt/confirm mock** — `askAnswer` 4 tests. **Requires test-name-keyed mock**: first test wants `confirm → true`, second `confirm → false`, third `prompt → "Alice"`, fourth `prompt → null`. Keyed via `_current-test-name` in the runner. Expected: +4.

29. **[done (+2) — partial, 4 parser-error tests remain (basic parse error messages, parse-error event, EOF newline crash, evaluate-api-first-error). All require stricter parser error-rejection — `add - to` currently parses silently to `(set! nil (hs-add-to! (- 0 nil) nil))`, `on click blargh end on mouseenter also_bad` parses silently to `(do (hs-on me "click" (fn (event) blargh)) (hs-on me "mouseenter" (fn (event) also_bad)))`. Plus emit-error-collection runtime + hyperscript:parse-error event with detail.errors. Larger than a single cluster budget; recommend bucket-D plan-first.] `hyperscript:before:init` / `:after:init` / `:parse-error` events** — 6 tests in `bootstrap` + `parser`. Fire DOM events at activation boundaries. Expected: +4-6.

30. **[done (+1)] `logAll` config** — 1 test. Global config that console.log's each command. Expected: +1.

### Bucket D: medium features (bigger commits, plan-first)

31. **[blocked: Bucket-D plan-first scope, doesn't fit one cluster budget. All 18 tests are SKIP (untranslated) — generator has no `error("HS")` helper. Required pieces: (a) generator-side `eval-hs-error` helper + recognizer for `expect(await error("HS")).toBe("MSG")` blocks; (b) runtime helpers `hs-null-error!` / `hs-named-target` / `hs-named-target-list` raising `'<sel>' is null`; (c) compiler patches at every target-position `(query SEL)` emit to wrap in named-target carrying the original selector source — that's ~17 command emit paths (add, remove, hide, show, measure, settle, trigger, send, set, default, increment, decrement, put, toggle, transition, append, take); (d) function-call null-check at bare `(name)`, `hs-method-call`, and `host-get` chains, deriving the leftmost-uncalled-name `'x'` / `'x.y'` from the parse tree; (e) possessive-base null-check (`set x's y to true` → `'x' is null`). Each piece is straightforward in isolation but the cross-cutting compiler change touches every emit path and needs a coordinated design pass. Recommend a dedicated design doc + multi-commit worktree like buckets E36-E40.] runtime null-safety error reporting** — 18 tests in `runtimeErrors`. When accessing `.foo` on nil, emit a structured error with position info. One coordinated fix in the compiler emit paths for property access, function calls, set/put. Expected: +15-18.

32. **[done (+7)] MutationObserver mock + `on mutation` dispatch** — 7 tests in `on`. Add MO mock to runner. Compile `on mutation [of attribute/childList/attribute-specific]`. Expected: +10-15.

33. **[done (+4) — partial, 1 test remains: `iterate cookies values work` needs `hs-for-each` to recognise host-array/proxy collections (currently `(list? collection)` returns false for the JS Proxy so the loop body never runs). Out of scope.] cookie API** — 5 tests in `expressions/cookies`. `document.cookie` mock in runner + `the cookies` + `set the xxx cookie` keywords. Expected: +5.

34. **[done (+7) — partial, 1 test remains: `every` keyword multi-handler-execute test needs handler-queue semantics where `wait for X` doesn't block subsequent invocations of the same handler — current `hs-on-every` shares the same dom-listen plumbing as `hs-on` and queues events implicitly via JS event loop, so the third synthetic click waits for the prior handler's `wait for customEvent` to settle. Out of single-cluster scope.] event modifier DSL** — 8 tests in `on`. `elsewhere`, `every`, `first click`, count filters (`once / twice / 3 times`, ranges), `from elsewhere`. Expected: +6-8.

35. **[done (+3)] namespaced `def`** — 3 tests. `def ns.foo() ...` creates `ns.foo`. Expected: +3.

### Bucket E: subsystems (DO NOT LOOP — human-driven)

All five have design docs on their own worktree branches pending review + merge. After merge, status flips to `design-ready` and they become eligible for the loop.

36. **[design-done, pending review — `plans/designs/e36-websocket.md` on `worktree-agent-a9daf73703f520257`] WebSocket + `socket`** — 16 tests. Upstream shape is `socket NAME URL [with timeout N] [on message [as JSON] …] end` with an **implicit `.rpc` Proxy** (ES6 Proxy lives in JS, not SX), not `with proxy { send, receive }` as this row previously claimed. Design doc has 8-commit checklist, +12–16 delta estimate. Ship only with intentional design review.

37. **[design-done, pending review — `plans/designs/e37-tokenizer-api.md` on `worktree-agent-a6bb61d59cc0be8b4`] Tokenizer-as-API** — 17 tests. Expose tokens as inspectable SX data via `hs-tokens-of` / `hs-stream-token` / `hs-token-type` etc; type-map current `hs-tokenize` output to upstream SCREAMING_SNAKE_CASE. 8-step checklist, +16–17 delta.

38. **[design-done, pending review — `plans/designs/e38-sourceinfo.md` on `agent-e38-sourceinfo`] SourceInfo API** — 4 tests. Inline span-wrapper strategy (not side-channel dict) with compiler-entry unwrap. 4-commit plan.

39. **[design-done, pending review — `plans/designs/e39-webworker.md` on `hs-design-e39-webworker`] WebWorker plugin** — 1 test. Parser-only stub that errors with a link to upstream docs; no runtime, no mock Worker class. Hand-write the test (don't patch the generator). Single commit.

40. **[design-done, pending review — `plans/designs/e40-real-fetch.md` on `worktree-agent-a94612a4283eaa5e0`] Fetch non-2xx / before-fetch event / real response object** — 7 tests. SX-dict Response wrapper `{:_hs-response :ok :status :url :_body :_json :_html}`; restructured `hs-fetch` that always fetches wrapper then converts by format; test-name-keyed `_fetchScripts`. 11-step checklist. Watch for regression on cluster-1 JSON unwrap.

### Bucket F: generator translation gaps (after bucket A-D)

Many tests are `SKIP (untranslated)` because `tests/playwright/generate-sx-tests.py` bailed with `return None`. These need patches to the generator to recognize more JS test patterns. Estimated ~25 recoverable tests. Defer to a dedicated generator-repair cluster once the queue above drains.

---

## Ground rules for the loop agent

1. **One cluster per commit.** Don't batch. Short commit message: `HS: <cluster name> (+N tests)`.
2. **Baseline first, verify at the end.** Before starting: record the current pass count for the target suite AND for one smoke range (0-195). After fixing: rerun both. Abort and mark blocked if:
   - Target suite didn't improve by at least +1.
   - Smoke range regressed (any test flipped pass → fail).
3. **Never edit `.sx` files with `Edit`/`Read`/`Write`.** Use sx-tree MCP (`sx_read_subtree`, `sx_replace_node`, `sx_insert_child`, `sx_insert_near`, `sx_replace_by_pattern`, `sx_rename_symbol`, `sx_validate`, `sx_write_file`).
4. **Sync WASM staging.** After every edit to `lib/hyperscript/<f>.sx`, run `cp lib/hyperscript/<f>.sx shared/static/wasm/sx/hs-<f>.sx`.
5. **Never edit `spec/tests/test-hyperscript-behavioral.sx` directly.** Fix the generator or the runtime.
6. **Scope:** `lib/hyperscript/**`, `shared/static/wasm/sx/hs-*`, `tests/hs-run-filtered.js`, `tests/playwright/generate-sx-tests.py`, `plans/hs-conformance-to-100.md`. Do not touch `spec/evaluator.sx`, the broader SX kernel, or unrelated files.
7. **Commit even partial fixes.** If you get +N where N is less than expected, commit what you have and mark the cluster `done (+N) — partial, <what's left>`.
8. **If stuck >30min on a cluster**, mark it `blocked (<reason>)` in the plan and move to the next pending cluster.
9. **Branch: `architecture`.** Commit locally. Never push. Never touch `main`.
10. **Log every iteration** in the Progress log below: one paragraph, what you touched, delta, commit SHA.
11. **Update the scoreboard** at `plans/hs-conformance-scoreboard.md` in the SAME plan-update commit: bump the `Merged:` line, update the row's `Status` / `Δ` / `Commit`, and adjust the buckets roll-up counts.
12. **Also expand scope to include** `plans/hs-conformance-scoreboard.md` (for rule 6 purposes).

## Known gotchas

- `env-bind!` creates bindings; `env-set!` mutates existing ones.
- SX `do` is R7RS iteration — use `begin` for multi-expr sequences.
- `cond` / `when` / `let` clause bodies evaluate only the last expr — wrap in `begin`.
- `list?` in SX checks for `{_type:'list'}` — it returns **false** on raw JS Arrays. `host-get node "children"` returns a JS Array in the mock, so recursion via `(list? kids)` silently drops nested elements.
- `append!` on a list-valued scoped var (`:s`) requires `emit-set` in the compiler — done, see commit 1613f551.
- When symbol target is `the-result`, also sync `it` (done, see emit-set).
- Hypertrace tests (196, 199, 200) and query-template test (615) hang under 200k step limit — always filter around them.
- `repeat forever` tests (1197, 1198) also hang.

## Progress log

(Reverse chronological — newest at top.)

### 2026-04-25 — Bucket F: in-expression filter semantics (+1)
- **67a5f137** — `HS: in-expression filter semantics (+1 test)`. `1 in [1, 2, 3]` was returning boolean `true` instead of the filtered list `(list 1)`. Root cause: `in?` compiled to `hs-contains?` which returns boolean for scalar items. Fix: (a) `runtime.sx` adds `hs-in?` returning filtered list for all cases, plus `hs-in-bool?` which wraps with `(not (hs-falsy? ...))` for boolean contexts; (b) `compiler.sx` changes `in?` clause to emit `(hs-in? collection item)` and adds new `in-bool?` clause emitting `(hs-in-bool? collection item)`; (c) `parser.sx` changes `is in` and `am in` comparison forms to produce `in-bool?` so those stay boolean. Suite hs-upstream-expressions/in: 8/9 → 9/9. Smoke 0-195: 173/195 unchanged.

### 2026-04-25 — cluster 22 window global fn fallback (+1)
- **d31565d5** — `HS cluster 22: simplify win-call emit + def→window + init-blocks test (+1)`. Two-part change building on 337c8265 (host-call-fn FFI + hs-win-call runtime). (a) `compiler.sx` removes the guard wrapper from bare-call and method-call `hs-win-call` emit paths — direct `(hs-win-call name (list args))` is sufficient since hs-win-call returns nil for unknown names; `def` compilation now also emits `(host-set! (host-global "window") name fn)` so every HS-defined function is reachable via window lookup. (b) `generate-sx-tests.py` fixes a quoting bug: `\"here\"` was being embedded as three SX nodes (`""` + symbol + `""`) instead of a single escaped-quote string; fixed with `\\\"` escaping. Hand-rolled deftest for `can refer to function in init blocks` now passes. Suite hs-upstream-core/regressions: 13/16 → 14/16. Smoke 0-195: 172/195 → 173/195.

### 2026-04-25 — cluster 11/33 followups: hide strategy + cookie clear (+2)
- **5ff2b706** — `HS: cluster 11/33 followups (+2 tests)`. Three orthogonal fixes that pick up tests now unblocked by earlier work. (a) `parser.sx` `parse-hide-cmd`/`parse-show-cmd`: added `on` to the keyword set that flips the implicit-`me` target. Previously `on click 1 hide on click 2 show` silently parsed as `(hs-hide! nil ...)` because `parse-expr` started consuming `on` and returned nil; now hide/show recognise a sibling feature and default to `me`. (b) `runtime.sx` `hs-method-call` fallback for non-built-in methods: SX-callables (lambdas) call via `apply`, JS-native functions (e.g. `cookies.clear`) dispatch via `(apply host-call (cons obj (cons method args)))` so the native receives the args list. (c) Generator `hs-cleanup!` body wrapped in `begin` (fn body evaluates only the last expr) and now resets `hs-set-default-hide-strategy! nil` + `hs-set-log-all! false` between tests — the prior `can set default to custom strategy` cluster-11 test had been leaking `_hs-default-hide-strategy` into the rest of the suite, breaking `hide element then show element retains original display`. New cluster-33 hand-roll for `basic clear cookie values work` exercises the method-call fallback. Suite hs-upstream-hide: 15/16 → 16/16. Suite hs-upstream-expressions/cookies: 3/5 → 4/5. Smoke 0-195 unchanged at 172/195.

### 2026-04-25 — cluster 35 namespaced def + script-tag globals (+3)
- **122053ed** — `HS: namespaced def + script-tag global functions (+3 tests)`. Two-part change: (a) `runtime.sx` `hs-method-call` gains a fallback for unknown methods — `(let ((fn-val (host-get obj method))) (if (callable? fn-val) (apply fn-val args) nil))`. This lets `utils.foo()` dispatch through `(host-get utils "foo")` when `utils` is an SX dict whose `foo` is an SX lambda. (b) Generator hand-rolls 3 deftests since the SX runtime has no `<script type='text/hyperscript'>` tag boot. For `is called synchronously` / `can call asynchronously`: `(eval-expr-cek (hs-to-sx (first (hs-parse (hs-tokenize "def foo() ... end")))))` registers the function in the global eval env (eval-expr-cek processes `(define foo (fn ...))` at top scope), then a click div is built via dom-set-attr + hs-boot-subtree!. For `functions can be namespaced`: define `utils` as a dict, register `__utils_foo` as a fresh-named global def, then `(host-set! utils "foo" __utils_foo)` populates the dict; click handler `call utils.foo()` compiles to `(hs-method-call utils "foo")` which now dispatches through the new runtime fallback. Skip-list cleared of the 3 def entries. Suite hs-upstream-def: 24/27 → 27/27. Smoke 0-195 unchanged at 172/195.

### 2026-04-25 — cluster 34 elsewhere / from-elsewhere modifier (+2)
- **3044a168** — `HS: elsewhere / from elsewhere modifier (+2 tests)`. Three-part change: (a) `parser.sx` `parse-on-feat` parses an optional `elsewhere` (or `from elsewhere`) modifier between event-name and source. The `from elsewhere` variant uses a one-token lookahead so plain `from #target` keeps parsing as a source expression. Emits `:elsewhere true` part. (b) `compiler.sx` `scan-on` threads `elsewhere?` (10th param) through every recursive call + new `:elsewhere` cond branch. The dispatch case becomes a 3-way `cond` over target: elsewhere → `(dom-body)` (listener attaches to body and bubble sees every click), source → from-source, default → `me`. The `compiled-body` build is wrapped with `(when (not (host-call me "contains" (host-get event "target"))) BODY)` so handlers fire only on outside-of-`me` clicks. (c) Generator drops `supports "elsewhere" modifier` and `supports "from elsewhere" modifier` from `SKIP_TEST_NAMES`. Suite hs-upstream-on: 48/70 → 50/70. Smoke 0-195 unchanged at 172/195.

### 2026-04-25 — cluster 34 count-filtered events + first modifier (+5 partial)
- **19c97989** — `HS: count-filtered events + first modifier (+5 tests)`. Three-part change: (a) `parser.sx` `parse-on-feat` accepts `first` keyword before event-name (sets `cnt-min/max=1`), then optionally parses a count expression after event-name: bare number = exact count, `N to M` = inclusive range, `N and on` = unbounded above. Number tokens coerced via `parse-number`. New parts entry `:count-filter {"min" N "max" M-or--1}`. (b) `compiler.sx` `scan-on` gains a 9th `count-filter-info` param threaded through every recursive call + a new `:count-filter` cond branch. The handler binding now wraps the `(fn (event) BODY)` in `(let ((__hs-count 0)) (fn (event) (begin (set! __hs-count (+ __hs-count 1)) (when COUNT-CHECK BODY))))` when count info is present. Each `on EVENT N ...` clause produces its own closure-captured counter, so `on click 1` / `on click 2` / `on click 3` fire on their respective Nth click (mix-ranges test). (c) Generator drops 5 entries from `SKIP_TEST_NAMES` — `can filter events based on count`/`...count range`/`...unbounded count range`/`can mix ranges`/`on first click fires only once`. Suite hs-upstream-on: 43/70 → 48/70. Smoke 0-195 unchanged at 172/195. Remaining cluster-34 work (`elsewhere`/`from elsewhere`/`every`-keyword multi-handler) is independent from count filters and would need a separate iteration.

### 2026-04-25 — cluster 29 hyperscript init events (+2 partial)
- **e01a3baa** — `HS: hyperscript:before:init / :after:init events (+2 tests)`. `integration.sx` `hs-activate!` now wraps the activation block in `(when (dom-dispatch el "hyperscript:before:init" nil) ...)` — `dom-dispatch` builds a CustomEvent with `bubbles:true`, the mock El's `cancelable` defaults to true, `dispatchEvent` returns `!ev.defaultPrevented`, so `when` skips the activate body if a listener called `preventDefault()`. After activation completes successfully it dispatches `hyperscript:after:init`. Generator (`tests/playwright/generate-sx-tests.py`) gains two hand-rolled deftests: `fires hyperscript:before:init and hyperscript:after:init` builds a wa container, attaches listeners that append to a captured `events` list, sets innerHTML to a div with `_=`, calls `hs-boot-subtree!`, asserts the events list. `hyperscript:before:init can cancel initialization` attaches a preventDefault listener and asserts `data-hyperscript-powered` is absent on the inner div after boot. Suite hs-upstream-core/bootstrap: 20/26 → 22/26. Smoke 0-195: 170 → 172. Remaining 4 cluster-29 tests (basic parse error messages, parse-error event, EOF newline, eval-API throws on first error) all need stricter parser error-rejection plus a parse-error collector — recommend bucket-D plan-first multi-commit, not a single iteration.

### 2026-04-25 — cluster 32 MutationObserver mock + on mutation dispatch (+7)
- **13e02542** — `HS: MutationObserver mock + on mutation dispatch (+7 tests)`. Five-part change: (a) `parser.sx` `parse-on-feat` now consumes `of <FILTER>` after `mutation` event-name. FILTER is one of `attributes`/`childList`/`characterData` (ident tokens) or one or more `@name` attr-tokens chained by `or`. Emits `:of-filter {"type" T "attrs" L?}` part. (b) `compiler.sx` `scan-on` threads new `of-filter-info` param; the dispatch case becomes a `cond` over `event-name` — for `"mutation"` it emits `(do on-call (hs-on-mutation-attach! target MODE ATTRS))` where ATTRS is `(cons 'list attr-list)` so the list survives compile→eval. (c) `runtime.sx` `hs-on-mutation-attach!` builds a config dict (`attributes`/`childList`/`characterData`/`subtree`/`attributeFilter`) matched to mode, constructs a real `MutationObserver(cb)`, calls `mo.observe(target, opts)`, and the cb dispatches a `"mutation"` event on target. (d) `tests/hs-run-filtered.js` replaces the no-op MO with `HsMutationObserver` (global registry, decodes SX-list `attributeFilter`); prototype hooks on `El.setAttribute/appendChild/removeChild/_setInnerHTML` fire matching observers synchronously, with `__hsMutationActive` re-entry guard so handlers that mutate the DOM don't infinite-loop. Per-test reset clears registry + flag. (e) `generate-sx-tests.py` drops 7 mutation entries from `SKIP_TEST_NAMES` and adds two body patterns: `evaluate(() => document.querySelector(SEL).setAttribute(N,V))` → `(dom-set-attr ...)`, and `evaluate(() => document.querySelector(SEL).appendChild(document.createElement(T)))` → `(dom-append … (dom-create-element …))`. Suite hs-upstream-on: 36/70 → 43/70. Smoke 0-195 unchanged at 170/195.

### 2026-04-25 — cluster 33 cookie API (partial +3)
- No `.sx` edits needed — `set cookies.foo to 'bar'` already compiles to `(dom-set-prop cookies "foo" "bar")` which becomes `(host-set! cookies "foo" "bar")` once the `dom` module is loaded, and `cookies.foo` becomes `(host-get cookies "foo")`. So a JS-only Proxy + Python generator change does the trick. Two parts: (a) `tests/hs-run-filtered.js` adds a per-test `__hsCookieStore` Map, a `globalThis.cookies` Proxy with `length`/`clear`/named-key get traps and a set trap that writes the store, and a `Object.defineProperty(document, 'cookie', …)` getter/setter that reads and writes the same store (so the upstream `length is 0` test's pre-clear loop over `document.cookie` works). Per-test reset clears the store. (b) `tests/playwright/generate-sx-tests.py` declares `(define cookies (host-global "cookies"))` in the test header and emits hand-rolled deftests for the three tractable tests (`basic set`, `update`, `length is 0`). Suite hs-upstream-expressions/cookies: 0/5 → 3/5. Smoke 0-195 unchanged at 170/195. Remaining `basic clear` and `iterate` tests need runtime.sx edits (hs-method-call fallback + hs-for-each host-array recognition) — out of scope for a JS-only iteration.

### 2026-04-25 — cluster 32 MutationObserver mock + on mutation dispatch (blocked)
- Two issues conspire: (1) `loops/hs` worktree has no pre-built sx-tree binary so MCP tools aren't loaded, and the block-sx-edit hook prevents raw `Edit`/`Read`/`Write` on `.sx` files. Built `hosts/ocaml/_build/default/bin/mcp_tree.exe` via `dune build` this iteration but tools don't surface mid-session. (2) Cluster scope is genuinely big: parser must learn `on mutation of <filter>` (currently drops body after `of` — verified via compile dump: `on mutation of attributes put "Mutated" into me` → `(hs-on me "mutation" (fn (event) nil))`), compiler needs `:of-filter` plumbing similar to intersection's `:having`, runtime needs `hs-on-mutation-attach!`, JS runner mock needs a real MutationObserver (currently no-op `class{observe(){}disconnect(){}}` at hs-run-filtered.js:348) plus `setAttribute`/`appendChild` instrumentation, and 7 entries removed from `SKIP_TEST_NAMES`. Recommended next step: dedicated worktree where sx-tree loads at session start, multi-commit shape (parser → compiler+attach → mock+runner → generator skip-list).

### 2026-04-25 — cluster 31 runtime null-safety error reporting (blocked)
- All 18 tests are `SKIP (untranslated)` — generator has no `error("HS")` helper at all. Inspected representative compile outputs: `add .foo to #doesntExist` → `(for-each ... (hs-query-all "#doesntExist"))` (silently no-ops on empty list, no error); `hide #doesntExist` → `(hs-hide! (hs-query-all "#doesntExist") "display")` (likewise); `put 'foo' into #doesntExist` → `(hs-set-inner-html! (hs-query-first "#doesntExist") "foo")` (passes nil through); `x()` → `(x)` (raises `Undefined symbol: x`, wrong format); `x.y.z()` → `(hs-method-call (host-get x "y") "z")`. Implementing this requires generator helper + 17 compiler emit-path patches + function-call/method-call/possessive-base null guards + new `hs-named-target`/`hs-named-target-list` runtime — too many surfaces for a single-iteration commit. Bucket D explicitly says "plan-first" — recommended path is a dedicated design doc and multi-commit worktree like E36-E40, not a loop iteration.

### 2026-04-24 — cluster 29 hyperscript:before:init / :after:init / :parse-error (blocked)
- **2b486976** — `HS-plan: mark cluster 29 blocked`. sx-tree MCP file ops returning `Yojson__Safe.Util.Type_error("Expected string, got null")` on every file-based call (sx_read_subtree, sx_find_all, sx_replace_by_pattern, sx_summarise, sx_pretty_print, sx_write_file). Only in-memory ops work (sx_eval, sx_build, sx_env). Without sx-tree I can't edit integration.sx to add before:init/after:init dispatch on hs-activate!. Investigated the 6 tests: 2 bootstrap (before/after init) need dispatchEvent wrapping activate; 4 parser tests require stricter parser error-rejection — `add - to` currently parses silently to `(set! nil (hs-add-to! (- 0 nil) nil))`, `on click blargh end on mouseenter also_bad` parses silently to `(do (hs-on me "click" (fn (event) blargh)) (hs-on me "mouseenter" (fn (event) also_bad)))`. Fundamental parser refactor is out of single-cluster budget regardless of sx-tree availability.

### 2026-04-24 — cluster 19 pick regex + indices
- **4be90bf2** — `HS: pick regex + indices (+13 tests)`. Parser: `pick items/item EXPR to EXPR` accepts `start`/`end` keywords; `pick match`/`pick matches` accept `| <flag>` after regex; `pick item N` without `to` still works. Runtime: `hs-pick-items`/`hs-pick-first`/`hs-pick-last` polymorphic for strings (slice) in addition to lists; `hs-pick-items` resolves `start`/`end` sentinel strings and negative indices at runtime; new `hs-pick-matches` wrapping `regex-find-all`; `hs-pick-regex-pattern` accepts `(list pat flags)` shape with `i` flag. Suite hs-upstream-pick: 11/24 → 24/24. Smoke 0-195 unchanged.

### 2026-04-24 — cluster 28 ask/answer + prompt/confirm mock
- **6c1da921** — `HS: ask/answer + prompt/confirm mock (+4 tests)`. Five-part change: (a) `tokenizer.sx` registers `ask` and `answer` as hs-keywords. (b) `parser.sx` — `cmd-kw?` gains both, `parse-cmd` gains cond branches dispatching to new `parse-ask-cmd` (emits `(ask MSG)`) and `parse-answer-cmd` which reads `answer MSG [with YES or NO]` with `parse-atom` for the choices (using `parse-expr` there collapsed `"Yes" or "No"` into `(or "Yes" "No")` before `match-kw "or"` could fire — parse-atom skips the logical layer). No-`with` form emits `(answer-alert MSG)`. (c) `compiler.sx` — three new cond branches (`ask`, `answer`, `answer-alert`) compile to `(let ((__hs-a (hs-XXX ...))) (begin (set! the-result __hs-a) (set! it __hs-a) __hs-a))` so `then put it into …` works. (d) `runtime.sx` — `hs-ask`, `hs-answer`, `hs-answer-alert` call `window.prompt`/`confirm`/`alert` via `host-call (host-global "window") …`. (e) `tests/hs-run-filtered.js` — test-name-keyed stubs for `globalThis.{alert,confirm,prompt}`, `__currentHsTestName` updated before each test. One extra tweak: `host-set!` innerHTML/textContent now coerces JS `null` → the string `"null"` (matching browser behaviour) so `prompt` returning null → `put it into #out` renders literal `"null"` text — the fourth test depends on exactly this. Suite hs-upstream-askAnswer: 1/5 → 5/5. Smoke 0-195: 166/195 → 170/195.

### 2026-04-24 — cluster 25 parenthesized commands and features
- **d7a88d85** — `HS: parenthesized commands and features (+1 test)`. Parser-only fix in `lib/hyperscript/parser.sx`. Three additions: (a) `parse-feat` gets a new cond branch — on `paren-open`, advance, recurse `parse-feat`, consume `paren-close`; lets features like `(on click ...)` be grouped. (b) `parse-cmd` gets two new cond branches — on `paren-close` return nil (so `cl-collect` terminates when the outer paren group ends), and on `paren-open` advance+recurse+close (parenthesized single commands like `(log me)`). (c) The key missing piece: `cl-collect` previously only recursed when the next token was a recognised command keyword (`cmd-kw?`), so after the first `(log me)` the next `(trigger foo)` would end the body. Extended the recursion predicate to also fire when the next token is `paren-open`. Result: `on click (log me) (trigger foo)` now emits both commands inside the handler body, not the second as a sibling top-level feature. Suite hs-upstream-core/parser: 9/14 → 10/14. Smoke 0-195: 165/195 → 166/195.

### 2026-04-24 — cluster 20 repeat property for-loops + where (worktree re-apply)
- **c932ad59** — `HS: repeat property for-loops + where (+3 tests)`. Worktree agent `a7c6dca2…` produced `c4241d57`; straight cherry-pick conflicted on runtime.sx with cluster 30's log-all block and cluster 27's intersection-attach helper, so logical diff was replayed surgically via sx-tree. Parser: `obj-collect` now `append`s pairs (not `cons`), preserving source order. Compiler: `emit-for` detects `coll-where` wrapping, binds the filter lambda to the for-loop variable name (not default `it`), and wraps symbol collections with `cek-try` instead of the broken `hs-safe-call` (uninitialised CEK call-ref in WASM). `array-index` emits `(hs-index …)` not `(nth …)`. Runtime: new polymorphic `hs-index` (dict/list/string/host dispatch); `hs-put-at!` default branch delegates to `hs-put!` when target is a DOM element; `hs-make-object` tracks insertion order in a hidden `_order` list; `hs-for-each` and `hs-coerce` (Keys/Entries/Map branches) prefer `_order` when present. Suite hs-upstream-repeat: 25/30 → 28/30 (remaining 2 are the `repeat forever` hangs we knowingly skip). Smoke 0-195 unchanged at 165.

### 2026-04-24 — cluster 27 intersection observer (worktree re-apply)
- **0c31dd27** — `HS: intersection observer mock + on intersection (+3 tests)`. Worktree agent `ad6e17cb…` produced the logical change but its commit's reformatted compiler/runtime bodies conflicted with cluster 26's landing of resize-observer state in `tests/hs-run-filtered.js` and `runtime.sx`; straight cherry-pick hit markers on `.sx` files that can't be Edit'd. Reapplied surgically via sx-tree: (a) parser `parse-on-feat` now collects `having margin X threshold Y` clauses between `from` and body, packaging them in `:having {"margin" M "threshold" T}`; (b) compiler `scan-on` threads a new `having-info` param and, when event-name is `"intersection"`, wraps `(hs-on target "intersection" fn)` with `(do on-call (hs-on-intersection-attach! target margin threshold))`; (c) runtime adds `hs-on-intersection-attach!` which constructs an `IntersectionObserver` with `{rootMargin, threshold}` opts and a callback that dispatches an `intersection` DOM event carrying `{intersecting, entry}` detail; (d) runner adds `HsIntersectionObserver` stub that fires the callback synchronously on `observe()` with `isIntersecting=true`. Suite hs-upstream-on: 33/70 → 36/70. Smoke 0-195 unchanged at 165.

### 2026-04-24 — cluster 26 resize observer (worktree agent cherry-pick)
- **304a52d2** (from worktree `worktree-agent-a8983e935d0d7a870` / `aea5f7d2`) — `HS: resize observer mock + on resize (+3 tests)`. `tests/hs-run-filtered.js`: `El.style` becomes a `Proxy` that fires a synthetic `resize` DOM event (detail carries `{width, height}` parsed from the current inline style) whenever `width`/`height` is written via direct property assignment OR `setProperty`. `ResizeObserver` no-op stub replaced by `HsResizeObserver` maintaining a per-element callback registry in `globalThis.__hsResizeRegistry`; added `ResizeObserverEntry` stub. `on resize` needs no parser/compiler work — `parse-compound-event-name` already accepts it and `hs-on` binds via `dom-listen`. Generator: new pattern translates `(page.)?evaluate(() => { document.(getElementById|querySelector)(...).style.PROP = 'VAL'; })` into `(host-set! (host-get TARGET "style") "PROP" "VAL")`. Suite hs-upstream-resize: 0/3 → 3/3. Smoke 0-195 unchanged at 165/195.

### 2026-04-24 — cluster 30 logAll config (worktree agent cherry-pick)
- **64bcefff** (from worktree `worktree-agent-a2bf303fd00e2fd4b` / `e50c3e6e`) — `HS: logAll config (+1 test)`. Runtime additions in `runtime.sx`: `_hs-config-log-all` flag + `_hs-log-captured` list + setters `hs-set-log-all!`, `hs-clear-log-captured!`, reader `hs-get-log-captured`, emitter `hs-log-event!` which both appends and forwards to `(host-call (host-global "console") "log" msg)`. `integration.sx` `hs-activate!` now emits `(hs-log-event! "hyperscript:init")` as the first action inside its when-block. Generator `tests/playwright/generate-sx-tests.py` detects the upstream body pattern (contains both `logAll` and `_hyperscript.config.logAll`) and emits a hand-rolled deftest: reset captured list → toggle log-all on → build detached `_="on click add .foo"` div → hs-boot-subtree! → assert `(some (fn (l) (string-contains? l "hyperscript:")) captured)`. Suite hs-upstream-core/bootstrap: 19/26 → 20/26. Smoke 0-195: 164 → 165.

### 2026-04-24 — cluster 24 properly interpolates values 2
- **cb37259d** — `HS-gen: string-aware line-comment stripping (+1 test)`. `process_hs_val` in `tests/playwright/generate-sx-tests.py` stripped `//…` line comments with a naïve regex, which devoured `https://yyy.xxxxxx.com/…` inside a backtick template — test 2074 was landing with the HS source truncated at `https:`. New helper `_strip_hs_line_comments` walks char-by-char and only strips `//` / leading-whitespace `--` when not inside `'…'`, `"…"`, or `` `…` ``; also respects `\\`-escapes inside strings. Regen produced full template intact. Suite hs-upstream-core/regressions: 11/16 → 12/16. Smoke 0-195: 163/195 → 164/195.

### 2026-04-24 — cluster 23 me symbol works in from expressions
- **0d38a75b** — `HS: closest parent <sel> traversal (+1 test)`. `parse-trav` now recognises `parent` as an ident modifier after the `closest` keyword: consumes it and re-invokes itself with kind `closest-parent`, so `closest parent <div/>` produces AST `(closest-parent "div" (me))` instead of `(string-postfix (closest "*" (me)) "parent")` — the latter was the generic trailing-ident-as-unit rule swallowing `parent`. Compiler translates `(closest-parent sel target)` to `(dom-closest (host-get target "parentElement") sel)` so `me` (the element with the `_` attribute) is skipped and only strict ancestors match. Also added `closest-parent` to the `put X into <trav>` inner-html shortcut alongside `next`/`previous`/`closest`. Suite hs-upstream-core/regressions: 10/16 → 11/16. Smoke 0-195: 162/195 → 163/195.

### 2026-04-24 — cluster 8 select returns selected text (cherry-picked from worktree)
- **0b9bbc7b** — `HS: select returns selected text (+1 test)`. Runtime `hs-get-selection` prefers `window.__test_selection` stash and falls back to `getSelection().toString()`. Compiler rewrites `(ref "selection")` to `(hs-get-selection)`. Generator detects `document.createRange() + getElementById(ID).firstChild + setStart/setEnd` and emits a single `host-set!` on `window.__test_selection` with the resolved substring, sidestepping a propagating DOM range/text-node mock. Runner resets `__test_selection` between tests. Suite hs-upstream-select: 3/4 → 4/4.

### 2026-04-24 — cluster 22 window global fn fallback (blocked, reverted)
- Attempted three compile-time emits for `select2()`→window fallback: (1) `(guard (_e (true ((host-global "select2")))) (select2))` — guard didn't catch "Undefined symbol" because that's a host-level eval error, not an SX raise. (2) `(if (env-has? "select2") (select2) ((host-global "select2")))` — `env-has?` primitive isn't loaded in the HS kernel (`Unhandled exception: "env-has?"`). (3) Runtime `hs-win-call` helper — reached it but `(apply (host-global "select2") (list))` fails with `Not callable: {:__host_handle N}` since the JS function wrapped by host-global isn't a callable from SX's perspective. Reverted all changes per abort rule. Proper fix: either expose `env-has?` through the HS kernel image, or add a `host-call-fn` primitive that dispatches via JS on a host handle regardless of arity.

### 2026-04-24 — cluster 21 possessive expression via its
- **f0c41278** — `HS: possessive expression via its (+1 test)`. Two generator changes: (a) `parse_run_locals` (Pattern 2 `var R = await run(...)`) now recognises `result: <literal>` in the opts dict and binds it to `it` so `run("its foo", {result: {foo: "foo"}})` produces `(eval-hs-locals "its foo" (list (list (quote it) {:foo "foo"})))`. Same extraction added to Pattern 1. (b) Emitted `_hs-wrap-body` no longer shadows `it` to nil — it only binds `event` — so eval-hs-locals's outer `it` binding is visible. `eval-hs` still binds `it` nil at its own fn wrapper. Suite hs-upstream-expressions/possessiveExpression: 22/23 → 23/23. Smoke 0-195: 162/195 unchanged.

### 2026-04-24 — cluster 18 throw respond via exception event
- **dda3becb** — `HS: throw respond via exception event (+2 tests)`. `hs-on` wraps each event handler in a `guard` that catches thrown exceptions and re-dispatches them as an `exception` DOM event on the same target with `{error: e}` as detail. `on exception(error)` handlers (also registered via hs-on) receive the event and destructure `error` from detail. Wrapping skips `exception`/`error` event handlers to avoid infinite loops. Suite hs-upstream-throw: 5/7 → 7/7. Smoke 0-195: 162/195 unchanged.

### 2026-04-24 — cluster 17 tell semantics (blocked, reverted)
- Attempted: drop the `me` shadow from `tell X` compile so `add .bar to me` preserves original and `put your innerText into me` writes to original. Fixed tests 2 and 3 but regressed 4 others (`restores a proper implicit me`, `works with an array`, `establishes a proper beingTold symbol`, `ignores null`) which require bare commands like `add .bar` (no explicit target) to use the told as default. Reverted per abort rule. Proper fix needs a distinct `beingTold` symbol with compiler rewriting bare commands to target `beingTold-or-me` while leaving explicit `me` alone — >30min cluster budget.

### 2026-04-24 — cluster 15 transition query-ref + multi-prop (partial +2)
- **3d352055** — `HS: transition query-ref + multi-prop (+2 tests)`. Three parts: (a) parser `collect-transitions` recognises `style` tokens (`*prop`) as a continuation, so `transition *width from A to B *height from A to B` chains both transitions instead of dropping the second. (b) Mock `El` class gets `nextSibling`/`previousSibling` (+ `*ElementSibling` aliases) so `transition *W of the next <span/>` can resolve the next-sibling target via host-get. (c) Generator pattern for `const X = await evaluate(() => { const el = document.querySelector(SEL); el.dispatchEvent(new Event(NAME, ...)); return ... })`; optionally prefixed by destructuring and allowing trailing `expect(...).toBe(...)` junk because `_body_statements` only splits on `;` at depth 0. Remaining `initial` test needs `on click N` count-filtered events. Suite hs-upstream-transition: 13/17 → 15/17. Smoke 0-195: 162/195 unchanged.

### 2026-04-24 — cluster 13 toggle multi-class + until (partial +2)
- **bd821c04** — `HS: toggle multi-class + until event (+2 tests)`. Parser `parse-toggle-cmd`: after the leading class ref collect any additional class refs and treat `toggle .foo .bar` as `toggle-between` (pair-only). Recognise `until EVENT [from SOURCE]` modifier and emit a new `toggle-class-until` AST. Compiler handles the new node by emitting `(begin (hs-toggle-class! tgt cls) (hs-wait-for src ev) (hs-toggle-class! tgt cls))` — reuses the cluster-9 event waiter so the class flips back when the event fires. `can toggle for a fixed amount of time` remains — sync mock io-sleep collapses the two toggles into one click frame; needs async scheduler. Suite hs-upstream-toggle: 22/25 → 24/25. Smoke 0-195: 162/195 unchanged.

### 2026-04-24 — cluster 12 show multi-element + display retention
- **98c957b3** — `HS: show multi-element + display retention (+2 tests)`. Two fixes in `tests/hs-run-filtered.js`: (a) `mt` (matches-selector) now splits comma-separated selector lists and matches if any clause matches, so `qsa("#d1, #d2")` returns both elements. (b) `host-get` on an `El` for `innerText` returns `textContent` (DOM-level alias) so `when its innerText contains "foo"` predicates can see the mock's stored text. Suite hs-upstream-show: 16/18 → 18/18. Smoke 0-195: 162/195 unchanged.

### 2026-04-24 — cluster 11 hide strategy (partial +3)
- **beb120ba** — `HS: hide strategy config (+3 tests)`. Three parts: (a) `runtime.sx` hs-hide-one!/hs-show-one! consult a new `_hs-hide-strategies` dict (and `_hs-default-hide-strategy` override) before falling through to the built-in display/opacity/hidden cases. Strategy fn is called directly with (op, el, arg). New setters `hs-set-hide-strategies!` and `hs-set-default-hide-strategy!`. (b) Generator `_hs_config_setup_ops` recognises `_hyperscript.config.defaultHideShowStrategy = "X"`, `delete …default…`, and `hideShowStrategies = { NAME: function (op, el, arg) { if … classList.add/remove } }` with brace-matched function body extraction. (c) Pre-setup emitter handles `__hs_config__` pseudo-name by emitting the SX expression as-is. Suite hs-upstream-hide: 12/16 → 15/16. Remaining test (`hide element then show element retains original display`) needs `on click N` count-filtered event handlers — separate feature. Smoke 0-195: 162/195 unchanged.

### 2026-04-24 — cluster 10 swap variable with property
- **30f33341** — `HS: swap variable with property (+1 test)`. Mock `El` class in `tests/hs-run-filtered.js`: `dataset` is now a `Proxy` that forwards property writes to `attributes["data-*"]`, and `setAttribute("data-*",...)` populates the backing dataset with camelCase key. That way `#target.dataset.val = "new"` updates the `data-val` attribute so the swap command can read+write the property correctly. Suite hs-upstream-swap: 3/4 → 4/4. Smoke 0-195: 162/195 unchanged.

### 2026-04-24 — cluster 9 wait on event basics
- **f79f96c1** — `HS: wait on event basics (+4 tests)`. Five parts: (a) `tests/hs-run-filtered.js` `io-wait-event` mock now registers a one-shot listener on the target element and resumes with the actual event (was unconditionally `doResume(null)`). (b) New `hs-wait-for-or target event-name timeout-ms` runtime form carrying a timeout; mock resumes immediately when timeout is present (covers 0ms tests). (c) `parser.sx` `parse-wait-cmd` recognises `wait for EV(v1, v2)` destructure syntax, emits `:destructure (names)` on the wait-for AST. (d) `compiler.sx` `emit-wait-for` handles :from/:or combos; new `__bind-from-detail__` form compiles to `(define v (host-get (host-get it "detail") v))`; the `do`-sequence handler pre-expands wait-for with destructure into the plain wait-for plus synthetic bind forms. (e) generator extracts `detail: ...` from CustomEvent option blocks. Suite `hs-upstream-wait`: 3/7 → 7/7. Smoke 0-195: 162/195 unchanged.
### 2026-04-23 — cluster 16 send can reference sender
- **ed8d71c9** — `HS: send can reference sender (+1 test)`. Three parts: (a) `emit-send` builds `{:sender me}` detail instead of nil for `send NAME target` and `send NAME`. (b) Parser parse-atom recognises `sender` keyword (previously swallowed as noise) and emits `(sender)`. (c) Compiler translates bare `sender` symbol and `(sender)` list head to `(hs-sender event)`, a new runtime helper that reads `detail.sender`. Suite hs-upstream-send: 7/8 → 8/8. Smoke 0-195: 162/195 unchanged.

### 2026-04-23 — cluster 14 unless modifier (blocked, reverted)
- Attempted: parser `cl-collect` handles `cmd unless cond` by emitting `(hs-unless-wrap cond cmd)`, compiler adds a `hs-unless-wrap` case that translates to `(if (hs-falsy? cond) cmd nil)`. Compile output correct. But test fails with `Undefined symbol: _test-result` suggesting the test-harness thunk eval throws somehow. Also added a generator pattern for `classList.add/remove/toggle` but that alone didn't fix the test. Reverted per abort rule; compile-shape fix looks sound and should be revisited after clusters that don't depend on harness internals.

### 2026-04-23 — cluster 8 select returns selected text (blocked, reverted)
- Attempted: added `hs-get-selection` runtime, compiler branch to rewrite bare `selection` to `(hs-get-selection)`, generator pattern to translate `evaluate(() => { var range = document.createRange(); ...; window.getSelection().addRange(range); })`, and mock support in `hs-run-filtered.js` for `document.createRange` / `window.getSelection` / `firstChild` text node. Tests still returned empty — range.toString() wasn't picking up the text. Reverted per the abort rule. Would need a more faithful mock of DOM text nodes with data propagation.

### 2026-04-23 — cluster 7 put hyperscript reprocessing (partial)
- **f21eb008** — `HS: put hyperscript reprocessing — generator fix (+1 test)`. Generator was swallowing non-window-setup `evaluate(() => { ... })` blocks. Fixed to only `continue` when a window-setup actually parsed, else fall through. Added a new pattern for `evaluate(() => { const e = new Event(...); SEL.dispatchEvent(e); })`. Suite hs-upstream-put: 33/38 → 34/38. "at end of" now passes; "at start of" / "in a element target" / "in a symbol write" still fail because the inserted-button handler doesn't activate on the afterbegin/innerHTML code paths. Smoke 0-195: 162/195 unchanged.

### 2026-04-23 — cluster 6 string template ${x}
- **108e25d4** — `HS: string template ${x} (+2 tests)`. Two-part fix: (a) `compiler.sx` now emits `(host-global "window")` (plus other well-known globals) for bare dot-chain base identifiers that would otherwise be unbound symbols. (b) `generate-sx-tests.py` now has `eval-hs-locals` ALSO call `host-set!` on `window.<name>` for each binding, so tests whose `window.X = Y` setup was translated as a local pair can still see `window.X`. Suite hs-upstream-expressions/strings: 5/8 → 7/8. Smoke 0-195: 162/195 unchanged.

### 2026-04-23 — cluster 5 some selector for nonempty match
- **e7b86264** — `HS: some selector for nonempty match (+1 test)`. `some <html/>` → `(not (hs-falsy? (hs-query-first "html")))` → `document.querySelector('html')`. Mock's querySelector searched from `_body`, missing the `_html` element. Fixed the mock to short-circuit for `html`/`body` and walk `documentElement`. Suite hs-upstream-expressions/some: 5/6 → 6/6. Smoke 0-195: 162/195 unchanged.

### 2026-04-23 — cluster 4 not precedence over or
- **4fe0b649** — `HS: not precedence over or + truthy/falsy coercion (+3 tests)`. `parse-atom`'s `not` branch emitted `(not (parse-expr))`, which let or/and capture the whole RHS, and also used SX's `not` which treats only nil/false as falsy. Fixed to emit `(hs-falsy? (parse-atom))` — tight binding + hyperscript truthiness (0, "", nil, false, []). Suite hs-upstream-expressions/not: 6/9 → 9/9. Smoke 0-195: 162/195 unchanged.

### 2026-04-23 — cluster 3 Values dict insertion order
- **e59c0b8e** — `HS: Values dict insertion order (+2 tests)`. Root cause was the OCaml kernel's dict implementation iterating keys in scrambled (non-insertion) order. Added `_order` hidden list tracked by `hs-values-absorb`, and taught `hs-coerce` FormEncoded/JSONString branches to iterate via `_order` when present (filtering the `_order` marker out). Suite hs-upstream-expressions/asExpression: 28/42 → 30/42. Smoke 0-195: 162/195 unchanged.

### 2026-04-23 — cluster 2 element→HTML via outerHTML
- **e195b5bd** — `HS: element → HTML via outerHTML (+1 test)`. Added an `outerHTML` getter on the mock `El` class in `tests/hs-run-filtered.js`. Merges `.id`/`.className` (host-set! targets) with `.attributes`, falls back to `innerText`/`textContent`. Suite hs-upstream-expressions/asExpression: 27/42 → 28/42. Smoke 0-195: 162/195 unchanged.

### 2026-04-23 — cluster 1 fetch JSON unwrap
- **39a597e9** — `HS: fetch JSON unwrap (+4 tests)`. Added `hs-host-to-sx` helper in `runtime.sx` that converts raw host-handle JS objects/arrays to proper SX dicts/lists via Object.keys/Array walks. `hs-fetch` now calls it on the result when format is `"json"`. Detects host-handle dicts by checking `(host-get v "_type") == "dict"` — genuine SX dicts have the marker, host handles don't. Suite hs-upstream-fetch: 11/23 → 15/23. Smoke 0-195: 162/195 unchanged.

### 2026-04-23 — cluster fixes session baseline
- **6b0334af** — `HS: remove bare @attr, set X @attr, JSON clean, FormEncoded, HTML join` (+3)
- **1613f551** — `HS add/append: Set dedup, @attr support, when-clause result tracking` (+6)
- Pre-loop baseline: 1213/1496 (81.1%).