refactor(parser): Plan 05b T8-A3 — the R1 remainder (U0-50, U0-12) by matthewevans · Pull Request #6719 · phase-rs/phase

matthewevans · 2026-07-28T04:17:28Z

Plan 05b tranche T8-A3 — the R1 remainder: U0-50 (Priority 14 ability-word "try as effect")
and U0-12 (CR 711.2a/711.2b LEVEL-block activated line). Follows #6716 (A2).

Three commits, one concern each. The two sites are asymmetric to an unusual degree: U0-50 stamps
nothing at all, and U0-12 has the only retry in phase A whose predicate runs on the lowered
definition.

U0-50 — the site whose shell stays `default()`

Entry point parse_effect_chain_with_context(&effect_text, AbilityKind::Spell, &mut ctx) — note
Spell, the one A3 site that is not Activated — so it takes parse_ability_ir_with_context, chosen
by name.

ir.shell stays AbilityShellIr::default(); no fields were invented for it. The conditional emit
guard stays at the call site: "should this be emitted at all" is control flow over the dispatch
loop, not a property of a definition, and folding it into the shell would make AbilityShellIr a
place where a recognizer can decline.

U0-12 — the retry whose predicate runs on the lowered def

The site parses, tests has_unimplemented(&def), and on failure re-parses a self-ref-normalized
variant through the same already-mutated ctx, keeping the alternative only if it comes back
clean. A0's H4 forbids an AbilityIr::from_definition, so this cannot lower-mutate-rewrap.

The shipped shape lowers purely to ask the predicate, while whichever IR won is what reaches
ability_ir_at. No IR-level has_unimplemented was invented — a second predicate over
EffectChainIr would be a rival authority free to diverge from the real one.

Three properties make it the same computation, each verified rather than assumed:

parse_effect_chain_with_context(t,k,cx) is lower_ability_ir(&parse_ability_ir_with_context(t,k,cx)),
so every has_unimplemented argument is bit-for-bit the definition the original tested.
ctx sequencing preserved. The retry's second parse receives the same already-mutated ctx,
not a fresh one, and the reset above it is not reordered. Interposing the lowering cannot perturb
this: lower_ability_ir takes no ParseContext, and a sweep for thread_local!/static mut/
AtomicU/RefCell</Cell< across parser/oracle_effect/ returns zero hits, so nothing on the
lowering path carries interior mutability. H4's non-idempotence concerns def → IR → def
round-trips, which do not exist here; applying a pure function twice is safe.
The predicate is invariant under the envelope — a property the plan did not state.
has_unimplemented reads only effect and sub_ability, both CR 608.2 resolution-tree fields, and
the shell stamps neither. So keeping the predicate pre-stamp is provably safe rather than merely
faithful.

Cost is one extra lowering per LEVEL-block activated line. That is intrinsic rather than laziness: the
predicate's lowered value is pre-shell and the emitted one is post-shell, so they are genuinely
different values and neither can be cached as the other. The path runs on nine cards.

The shape is also already precedented in the tree — the instant/sorcery prevention-text site does
has_unimplemented(&lower_effect_chain_ir(&effect_ir)) around a retained OracleNodeIr::Spell.

Restrictions: parsed constraints.restrictions lead, then LevelCounterRange{minimum,maximum} —
same polarity as A2. Stages [ExtractCostReduction, ExtractManaSpendTrigger]. The triggered branch
below, whose And graft is deliberately post-lowering, is untouched.

§5.3 — reachability first, and the silences are distinguished

Reachability probes attributed by message, never by name — which mattered, since neither U0-50
reacher's name mentions the recognizer:

site	`--lib`	integration
U0-50	`swallow_check::tests::duration_until_eot_drop_tower`	`printed_ability_order::memory_test_triggers_are_in_printed_order`
U0-12	`oracle::tests::leveler_activated_abilities_get_level_counter_range`	`printed_ability_order::under_construction_skyscraper_abilities_are_in_printed_order`

Exactly one test per binary per site. Because a panic! at block entry fails every test that enters,
that enumeration is complete, which licenses scoping perturbations to those tests by argument
rather than by hope.

Perturbations ran through a harness carrying a mandatory-red liveness arm, so a dead harness could
not read as "all silent". Liveness → red, control → green, in both binaries.

perturbation	result
drop `LevelCounterRange`	RED — the only witnessed axis
swap parsed-vs-implicit order	silent
drop `ExtractCostReduction` / drop `ExtractManaSpendTrigger` / reorder stages	silent
retry never fires	silent
`ChainLoweringMode` → `Standalone`	silent (null reported as null)
U0-50 emit guard inverted	silent across both full suites

The silences were then diagnosed rather than lumped together, over the full nine-card leveler
population:

Restriction order is UNREACHABLE, not merely untested. constraints.restrictions is empty for
every leveler card, so the composed vec is always one element and a reversal is a no-op. There is
no non-degenerate pool card, so the A1/A2 remediation recipe is impossible here rather than
skipped.
The retry is dead on the pool, and disjointly so. Its gate opens for exactly one card —
Under-Construction Skyscraper, whose Add {W}, {B}, {G}, or {C} lowers to Unimplemented — but that
text has no self-reference, so normalized_effect != effect_text is false and the alt parse never
runs. Every card that does self-reference parses cleanly and never opens the gate. The most
intricate logic in the tranche is unwitnessed and currently unreachable.
ExtractManaSpendTrigger's guard is LIVE here — the first time in phase A. A2 proved zero pool
cards at its four sites lower to a root Effect::Mana; Joraga Treespeaker's {T}: Add {G}{G} does.
The fold body remains inert (it additionally needs a trailing "when you spend this mana" node, and
no leveler prints one), so dropping the fold is correctly silent. Stated as guard-live /
body-dead rather than as a blanket "first live witness".

Remediation

Additive two-layer snapshots for Guul Draz Assassin (two striations, two-component {B}, {T}
cost, bounded 2-3 plus unbounded 4+) and Joraga Treespeaker (pins the live mana root), Oracle
text verbatim from the card pool. Both watched go red under a dropped LevelCounterRange, then
restored.

Gates

cargo fmt --all                                      clean
cargo clippy -p engine --lib -- -D warnings          zero warnings
cargo nextest run -p engine --lib                    17851 passed, 6 skipped   (17849 + 2 new)
cargo nextest run -p engine --test integration       4123 passed, 2 skipped
./scripts/check-prelowered-ratchet.sh                Gate P PASS
./scripts/check-skill-doc.sh                         ✓ oracle-parser skill references valid
./scripts/check-parser-combinators.sh                Gate G PASS / Gate A PASS

Zero existing snapshots changed: git diff --name-status is 4 A / 2 M, both modified files
source. .snap.new count 0. oracle.rs unmoved at 16 on the ratchet; ledger unedited.

Parser string-dispatch diff gate clean, including the rfind/split/split_once/rsplit
blind-spot arms. CR diff gate: 8 numbers, 0 unverified, text read for each — 106.6, 601.2f, 602.1a,
602.1b, 603.3, 608.2, 711.2a, 711.2b.

Harvest (§5.4)

Kargan Dragonlord's LEVEL 8+ firebreathing is silently swallowed — pre-existing, and a class
bug. parse_level_blocks (oracle_level.rs) attempts static classification before
activated-colon detection, and {R}: this creature gets +1/+0 until end of turn. matches a static
line, which consumes it with a continue. The card yields zero abilities where all eight
sibling levelers yield gated ones — and an unconsumed line would instead fall through and yield
an ungated ability, so the observed zero pins the static branch as the consumer. Net effect: a
repeatable {R}-per-activation pump becomes a permanent static bonus (CR 711.2 + CR 602.1). Any
level-block activated ability whose post-colon text parses as a static is affected.
Under-Construction Skyscraper's Add {W}, {B}, {G}, or {C} is Unimplemented at both
striations — a real parser gap for "add one of N colors" mana abilities.
duration_until_eot_drop_tower is a vacuous negative. Its only assertion is
!has_swallowed_detector(…), which holds when the ability is never emitted at all — which is
precisely why inverting U0-50's emit guard is silent.
Restriction-order at U0-12 is unreachable by construction on the current pool.
The retry is dead on the current pool by disjointness of its two conditions.

One limitation stated rather than papered over

U0-50's emit guard polarity is unwitnessed. A Drop Tower fixture was built, measured, and
discarded: modelling the card correctly — Artifact plus its verbatim Attraction subtype —
routes the Visit line to the CR 717 preprocessor, and a re-inserted probe confirmed the site is never
reached. The two in-suite reachers hit U0-50 only because they supply no subtype, i.e. only because
they model the card wrongly. Shipping that fixture would have "closed" the gap with a test that passes
for the wrong reason.

For the record, this does not put the site's production reachability in doubt: U0-50 is the general
ability-word handler (its own comment cites "Void Shields —"), not an Attraction path, and 1,027
cards in the pool carry an ability-word prefix line. The unwitnessed axis is the guard's polarity,
not the site.

Summary by CodeRabbit

Bug Fixes
- Improved parsing of level-based and ability-word effects, including activation costs, descriptions, and level restrictions.
- Better handling of self-referential ability text and previously unsupported content.
Tests
- Added coverage for level-based activated abilities and level-based mana abilities.
- Added snapshots validating activation restrictions and mana ability details.

…ys default Priority 14 ability-word "try as effect" (census row 170). The only site in the family that stamps **nothing** — not `cost`, not `activation_restrictions`, not even `description` — so `ir.shell` stays `AbilityShellIr::default()` and the conversion is the bare entry-point swap with no envelope to carry and no `ShellStage` to list. No fields were invented for it. **The conditional emit stays at the call site, on purpose.** "Should this be emitted at all" is control flow over the enclosing dispatch loop, not a property of the definition being built; folding it into the shell would make `AbilityShellIr` a place where a recognizer can decline, which is not what the CR 602.1 envelope describes. The guard is therefore untouched. **How the lowered-def predicate survives the conversion.** `has_unimplemented` reads an `AbilityDefinition`, and an `AbilityDefinition` cannot be un-lowered into an `AbilityIr` (there is no `AbilityIr::from_definition`, by design), so the predicate cannot move onto the IR. It instead runs on `lower_ability_ir(&ir)` while the *retained* artifact stays the IR. This is not a new idiom: the instant/sorcery prevention-text site a few hundred lines above already routes `has_unimplemented(&lower_effect_chain_ir(&effect_ir))` around a retained `OracleNodeIr::Spell`. Lowering twice (once for the predicate, once inside `ability_ir_at`) is a repeat of the same computation, never a different one: `lower_ability_ir` is a pure `&AbilityIr -> AbilityDefinition` — it takes no `ParseContext`, and `rg 'thread_local!|static mut|AtomicU|RefCell<|Cell<' crates/engine/src/parser/oracle_effect/` returns zero hits, so nothing under the whole of the lowering path carries interior mutability. H4's non-idempotence is about `def -> IR -> def` round-tripping, which is exactly what does not exist; it says nothing about applying a pure function twice to the same input. Mode inherited by NAME: the original called `parse_effect_chain_with_context`, so this takes `parse_ability_ir_with_context`. `AbilityKind::Spell` is preserved — this site is the one in A3 that is not `Activated`. CR-faithfulness: **by construction**, for the whole class of text the recognizer handles; no property of any card's text participates. `parse_effect_chain_with_context(t,k,cx)` IS `lower_ability_ir(&parse_ability_ir_with_context(t,k,cx))` (`oracle_effect/mod.rs`), and phase-A `ability_ir_at(l, ir)` IS `ability_at(l, lower_ability_ir(&ir))` (`oracle.rs`). With a `default()` shell and no stages the envelope applier is a no-op, so the converted site computes exactly the old site's value. Reachability (probe-attributed, never by test name): `panic!` at the site fires in `parser::swallow_check::tests::duration_until_eot_drop_tower` (--lib) and `printed_ability_order::memory_test_triggers_are_in_printed_order` (--test integration). Note neither name mentions this recognizer — attribution by probe message was load-bearing here, exactly as A2 warned. Gates: fmt clean, `Gate P PASS` (oracle.rs unmoved at 16), skill-doc `✓`, Gate G + Gate A PASS, `cargo clippy -p engine --lib -D warnings` clean, and zero snapshot churn (no `*.snap.new`; `git status` shows only this file).

…y runs on the lowered def CR 711.2a + CR 711.2b activated line inside a LEVEL block (census row 132), and the only site in phase A whose *retry decision* is made on a LOWERED definition while the *retained artifact* must stay an IR. **The retry, and why this shape is behavior-identical.** `AbilityIr` has no `from_definition` and `lower_ability_ir` is not idempotent, so an `AbilityDefinition` cannot be un-lowered — the retry cannot be expressed by lowering, mutating and re-wrapping. Both candidates are therefore parsed as IR and each is lowered purely to *ask* `has_unimplemented`, while whichever IR won is what reaches `ability_ir_at`. No IR-level `has_unimplemented` was invented: a second predicate over `EffectChainIr` would be a rival authority free to diverge from the real one, so this lowers and asks the real one. Three properties make it the same computation as the original: 1. `parse_effect_chain_with_context(t,k,cx)` IS `lower_ability_ir(&parse_ability_ir_with_context(t,k,cx))`, so each `has_unimplemented` argument is bit-for-bit the definition the original tested. 2. **The `ctx` sequencing is preserved exactly.** The retry's second parse receives the SAME, already-mutated `ctx` as the first — not a fresh one — and the `ctx` reset above it is not reordered. Interposing the lowering between the two parses cannot perturb this: `lower_ability_ir` takes no `ParseContext`, and `rg 'thread_local!|static mut|AtomicU|RefCell<|Cell<' crates/engine/src/parser/oracle_effect/` returns zero hits, so nothing on the lowering path carries interior mutability. `lower_ability_ir` is a pure `&AbilityIr -> AbilityDefinition` and calling it twice is safe — verified by reading it, not assumed. 3. The predicate is invariant under the envelope: `has_unimplemented` reads only `effect` and `sub_ability`, both CR 608.2 resolution-tree fields, and the shell stamps neither. So keeping the predicate pre-stamp (where it already was) is safe rather than merely convenient. Cost: one extra lowering per LEVEL-block activated line (two or three rather than one or two). It is intrinsic, not laziness — the predicate's lowered value is *pre*-shell and the emitted one is *post*-shell, so they are genuinely different values and neither can be cached as the other. The path runs only on LEVEL blocks: nine cards in today's pool. Composed order — restrictions: parsed `constraints.restrictions` LEAD, then the implicit `LevelCounterRange{minimum,maximum}` trails. Same polarity as A2's Boast/Exhaust/Forecast, opposite of A1's U0-22. Stages: `[ExtractCostReduction, ExtractManaSpendTrigger]`, the order the site ran the two folds in; `lower_ability_ir` runs stages after the envelope stamps, which is the pinned order the site had by hand. The original's `=` is equivalent to the shell's `extend`, verified not assumed: `rg activation_restrictions crates/engine/src/parser/oracle_effect/` returns 3 hits, all inside `apply_ability_shell_envelope` itself, so nothing reachable from `lower_ability_ir` writes the root's restrictions and the field is empty when the shell runs. `cost` still comes from `normalize_self_refs_for_static(cost_text, card_name)` before the colon split — that normalization is untouched. `description` is the full `ability_text`, not `effect_text`. The triggered-ability branch below, whose `TriggerCondition::And` graft must stay post-lowering to keep `And[gate,x,y]` flat, is deliberately NOT touched. Mode inherited by NAME: the original called `parse_effect_chain_with_context` twice, so both calls take `parse_ability_ir_with_context`. CR-faithfulness: **by construction**, for the whole class of text the recognizer handles; no property of any card's text participates. CR numbers grep-verified against docs/MagicCompRules.txt with the rule text read, not just the number matched: 711.2 (:5798), 711.2a (:5800), 711.2b (:5802) — the level striation gating the printed abilities; 602.1a (:2516) "everything before the colon"; 602.1b (:2519) activation instructions restricting when a player may activate. Gates: fmt clean, `Gate P PASS` (oracle.rs unmoved at 16), skill-doc OK, Gate G + Gate A PASS, `cargo clippy -p engine --lib -D warnings` clean, `cargo nextest run -p engine --lib` 17849 passed / 6 skipped, `--test integration` 4123 passed / 2 skipped, zero snapshot churn (no `*.snap.new`; only this file modified).

…e (§5.3 remediation) Additive two-layer snapshots. No production code and no pre-existing snapshot is touched: `git diff --name-status` is one `M` on `snapshot_tests.rs` and four `A`s. **What was unwitnessed at U0-12.** The only pre-existing test over the site (`oracle_tests::leveler_activated_abilities_get_level_counter_range`) asserts `LevelCounterRange` presence via `.contains(…)` — order-insensitive — and asserts nothing about the `cost` or `description` the site also stamps. Of the perturbations run against it, only *dropping* `LevelCounterRange` went red. * **Guul Draz Assassin** — the richest fixture in the nine-card leveler population: two striations, a two-component `{B}, {T}` cost (so the CR 602.1a stamp is pinned to more than a bare `{T}`), a targeted effect, and two different range shapes (bounded `2-3`, unbounded `4+`) so a range collapsing to a constant would show. * **Joraga Treespeaker** — pins the first LIVE `ExtractManaSpendTrigger` guard in phase A. A2 established its four keyword sites can never run that fold (no pool card with those keywords lowers to a root `Effect::Mana`); Joraga's `LEVEL 1-4` body `{T}: Add {G}{G}.` does, so the guard passes here for the first time. The fold's *body* is still inert — it additionally needs a trailing "when you spend this mana …" sub-ability and no leveler prints one — so this fixture is the one that starts discriminating the moment a striation prints a spend trigger. Both fixtures are pool-verified, not synthetic: Oracle text, types and subtypes are verbatim from `data/card-data.json`. Both were **watched go red** — dropping the `LevelCounterRange` push fails both, and reverting it restores green. **U0-50 is deliberately NOT remediated here, and that is a finding.** Inverting its emit guard left both full suites green. Exactly two tests reach it and neither constrains it: `duration_until_eot_drop_tower` asserts only `!has_swallowed_detector(…)`, a negative that holds vacuously when nothing is emitted, and `memory_test_triggers_are_in_printed_order` constrains only triggers. A snapshot fixture was built for Drop Tower and then DISCARDED after measurement: modelling the card correctly — `Artifact` with its verbatim `Attraction` subtype — routes the Visit line to the CR 717 preprocessor and the site is never reached. The two in-suite reachers hit U0-50 only because they supply no subtype, i.e. only because they model the card wrongly. So no correctly-modelled fixture can currently witness that guard, and identifying U0-50's true production population is a separate question, raised to the lead rather than papered over with a fixture that passes for the wrong reason. Gates: fmt clean, Gate P PASS, skill-doc OK, Gate G + Gate A PASS, `cargo clippy -p engine --lib -D warnings` clean, `cargo nextest run -p engine --lib` 17851 passed / 6 skipped (two new), zero `*.snap.new`, zero pre-existing snapshots modified.

coderabbitai · 2026-07-28T04:17:56Z

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 4176c245-6e00-433b-8c6e-23c604c4e4fb

📥 Commits

Reviewing files that changed from the base of the PR and between e001db0 and cfffef8.

⛔ Files ignored due to path filters (4)

crates/engine/src/parser/oracle_ir/snapshots/engine__parser__oracle_ir__snapshot_tests__guul_draz_assassin_ir.snap is excluded by !**/*.snap, !**/snapshots/**
crates/engine/src/parser/oracle_ir/snapshots/engine__parser__oracle_ir__snapshot_tests__guul_draz_assassin_lowered.snap is excluded by !**/*.snap, !**/snapshots/**
crates/engine/src/parser/oracle_ir/snapshots/engine__parser__oracle_ir__snapshot_tests__joraga_treespeaker_ir.snap is excluded by !**/*.snap, !**/snapshots/**
crates/engine/src/parser/oracle_ir/snapshots/engine__parser__oracle_ir__snapshot_tests__joraga_treespeaker_lowered.snap is excluded by !**/*.snap, !**/snapshots/**

📒 Files selected for processing (2)

crates/engine/src/parser/oracle.rs
crates/engine/src/parser/oracle_ir/snapshot_tests.rs

📝 Walkthrough

Walkthrough

Activated ability parsing now uses AbilityIr through validation, staged extraction, metadata assignment, and emission. New snapshot fixtures cover leveler activated abilities and a mana-producing level ability.

Changes

Activated ability IR parsing

Layer / File(s)	Summary
IR parsing and staged emission `crates/engine/src/parser/oracle.rs`	LEVEL-block and ability-word effect branches parse and validate `AbilityIr`, apply shell metadata and extraction stages, and emit through `ability_ir_at`.
Leveler snapshot coverage `crates/engine/src/parser/oracle_ir/snapshot_tests.rs`	Adds snapshots for Guul Draz Assassin level activations and Joraga Treespeaker’s mana-producing level ability.

Estimated code review effort: 3 (Moderate) | ~25 minutes

Sequence Diagram(s)

sequenceDiagram
  participant OracleParser
  participant AbilityIrLowerer
  participant AbilityEmitter
  OracleParser->>OracleParser: Parse activated ability into AbilityIr
  OracleParser->>AbilityIrLowerer: Check lowered IR for unimplemented content
  OracleParser->>AbilityEmitter: Emit with ability_ir_at

Possibly related PRs

phase-rs/phase#6699: Updates adjacent parse_oracle_ir LEVEL-block IR emission behavior.
phase-rs/phase#6714: Introduces the shell widening and staged lowering infrastructure used here.
phase-rs/phase#6715: Applies the related activated-ability AbilityIr parsing and emission pattern.

Suggested labels: enhancement, quality

Suggested reviewers: lgray, dripsmvcp, invalidcards

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title matches the parser refactor and the U0-50/U0-12 areas changed in the PR.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

Create stacked PR
Commit on current branch

🧪 Generate unit tests (beta)

Create PR with unit tests
Commit unit tests in branch ship/plan05b-t8a3

_{Comment @coderabbitai help to get the list of available commands.}

github-actions · 2026-07-28T04:32:34Z

Parse changes introduced by this PR

✓ No card-parse changes detected.

matthewevans added 3 commits July 27, 2026 21:15

matthewevans enabled auto-merge July 28, 2026 04:17

matthewevans added this pull request to the merge queue Jul 28, 2026

Merged via the queue into main with commit f546654 Jul 28, 2026
15 checks passed

matthewevans deleted the ship/plan05b-t8a3 branch July 28, 2026 04:48

matthewevans mentioned this pull request Jul 28, 2026

refactor(parser): Plan 05b T8-A4 — U0-53, the one R2 row under the cap #6720

Merged

coderabbitai Bot mentioned this pull request Jul 28, 2026

refactor(parser): Plan 05b — the shared residual node and optional on the ability shell #6733

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

refactor(parser): Plan 05b T8-A3 — the R1 remainder (U0-50, U0-12) - #6719

refactor(parser): Plan 05b T8-A3 — the R1 remainder (U0-50, U0-12)#6719
matthewevans merged 3 commits into
mainfrom
ship/plan05b-t8a3

matthewevans commented Jul 28, 2026 •

edited by coderabbitai Bot

Loading

Uh oh!

coderabbitai Bot commented Jul 28, 2026 •

edited

Loading

Walkthrough

Changes

Sequence Diagram(s)

Uh oh!

github-actions Bot commented Jul 28, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Uh oh!

Conversation

matthewevans commented Jul 28, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

U0-50 — the site whose shell stays default()

U0-12 — the retry whose predicate runs on the lowered def

§5.3 — reachability first, and the silences are distinguished

Remediation

Gates

Harvest (§5.4)

One limitation stated rather than papered over

Summary by CodeRabbit

Uh oh!

coderabbitai Bot commented Jul 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Sequence Diagram(s)

Uh oh!

github-actions Bot commented Jul 28, 2026

Parse changes introduced by this PR

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

matthewevans commented Jul 28, 2026 •

edited by coderabbitai Bot

Loading

U0-50 — the site whose shell stays `default()`

coderabbitai Bot commented Jul 28, 2026 •

edited

Loading