Recommended actions

• [Documentation problem | High] Clarify named-index syntax in the Rust server skill — The benchmark shows repeated misuse of #[index(...)] as a field attribute when the expected form is table-level index(accessor = ..., btree(columns = [...])). Evidence: t_018_constraints, t_051_denormalized_index.
• [Documentation problem | High] Update the scheduled-table examples to match the expected schema and reducer signatures — The benchmark expected scheduled tables without the extra public table flag, with non-pub fields, and with lifecycle reducers returning Result<(), String>. Evidence: t_002_scheduled_table, t_038_schedule_at_time.

Failure patterns

Named index annotations were placed on fields instead of on the table (2 tasks)

• Classification: Documentation problem
• Confidence: High
• What the model wrote:
  • #[index(btree, accessor = by_name)] pub name: String,
  • #[index(btree, accessor = by_category)] pub category_id: u64,
  • #[index(btree, accessor = by_category_slug)] pub category_slug: String,
• What was expected:
  • #[table(..., index(accessor = by_name, btree(columns = [name])))]
  • #[table(..., index(accessor = by_category, btree(columns = [category_id])))]
  • #[table(..., index(accessor = by_category_slug, btree(columns = [category_slug])))]
• What the error says: publish_error: spacetime publish failed (exit=1)
• Why this happened: The supplied guidance does not make the named-index form distinct enough from inline single-column indexing, so the model used an unsupported field-level syntax with accessor = ....
• Suggested action: Update the Rust server skill to show named indexes only in table-level attributes, with a side-by-side example contrasting inline single-column indexing and named index declarations.
• Suggested area: rust-server/SKILL.md
• Affected tasks: t_018_constraints, t_051_denormalized_index
• AI guidelines gap: The guidance does not explicitly state that accessor = ... requires table-level index(...) syntax.

Scheduled-table syntax and lifecycle shape diverged from the benchmark’s expected form (2 tasks)

• Classification: Documentation problem
• Confidence: High
• What the model wrote:
  • #[table(accessor = tick_timer, scheduled(tick), public)]
  • pub scheduled_at: ScheduleAt,
  • #[reducer] pub fn tick(_ctx: &ReducerContext, _timer: TickTimer) { }
  • #[reducer(init)] pub fn init(ctx: &ReducerContext) { ... }
• What was expected:
  • #[table(accessor = tick_timer, scheduled(tick))]
  • scheduled_id: u64, / scheduled_at: ScheduleAt, without pub
  • pub fn tick(_ctx: &ReducerContext, _row: TickTimer) -> Result<(), String> { Ok(()) }
  • pub fn init(ctx: &ReducerContext) -> Result<(), String> { ... }
• What the error says: schema_parity: failed
• Why this happened: The scheduled-table section in the supplied guidance does not match the benchmark’s current expected schema shape, especially around table visibility, field visibility, and reducer return types.
• Suggested action: Update the scheduled-table examples and reducer signatures in the Rust server skill to match the current accepted schema exactly.
• Suggested area: rust-server/SKILL.md
• Affected tasks: t_002_scheduled_table, t_038_schedule_at_time
• AI guidelines gap: The guidance does not clearly distinguish the benchmark’s required scheduled-table declaration from the more general table examples.

Embedded type and table metadata did not match the expected minimal schema (1 task)

• Classification: Model limitation
• Confidence: Low
• What the model wrote:
  • #[derive(SpacetimeType)]
  • #[spacetimedb::table(accessor = entity, public)]
• What was expected:
  • #[derive(SpacetimeType, Clone, Debug)]
  • #[table(accessor = entity)]
• What the error says: schema_parity: failed
• Why this happened: The model overfit to the provided example style and did not reproduce the exact schema and derives used by the benchmark.
• Suggested action: None
• Suggested area: None
• Affected tasks: t_003_struct_in_table
• AI guidelines gap: None

Right-semijoin operands were reversed (1 task)

• Classification: Model limitation
• Confidence: Medium
• What the model wrote:
  • ctx.from.member().right_semijoin(
  • ctx.from.eligibility(),
  • |member, eligibility| member.id.eq(eligibility.member_id),
• What was expected:
  • ctx.from.eligibility().right_semijoin(ctx.from.member(), |eligibility, member| { ... })
• What the error says: semijoin_intersection: failed
• Why this happened: The model inverted the left/right side of right_semijoin and supplied the closure arguments in the wrong order.
• Suggested action: None
• Suggested area: None
• Affected tasks: t_062_semijoin_intersection
• AI guidelines gap: None

Timestamp-window view used the wrong constructor/API shape (1 task)

• Classification: Model limitation
• Confidence: Medium
• What the model wrote:
  • Timestamp::from_micros_since_unix_epoch(200).unwrap();
  • let mut rows: Vec<Event> = ...collect();
  • rows.sort_by(...)
• What was expected:
  • Timestamp::from_micros_since_unix_epoch(200);
  • ctx.db.event().occurred_at().filter(start..=end).collect()
• What the error says: schema_parity: failed
• Why this happened: The model assumed the timestamp constructor returned an Option and added unnecessary post-processing instead of using the direct range query shown in the golden answer.
• Suggested action: None
• Suggested area: None
• Affected tasks: t_063_timestamp_window
• AI guidelines gap: None

External upload flow missed the router and body-handling details (1 task)

• Classification: Model limitation
• Confidence: Medium
• What the model wrote:
  • No #[router] fn routes() -> Router { ... }
  • Body::from_bytes("https://files.local/object-1".to_string())
  • pub fn upload_and_register(...) -> String
• What was expected:
  • #[router] fn routes() -> Router { Router::new().post("/upload", upload) }
  • Body::from_bytes("https://files.local/object-1")
  • The procedure body that clones the upload URL, checks response.status().is_success(), and stores the uploaded asset in a transaction
• What the error says: publish_error: spacetime publish failed (exit=1)
• Why this happened: The model mixed procedure, router, and HTTP-handler patterns and did not follow the expected outbound HTTP/body conventions from the benchmark.
• Suggested action: None
• Suggested area: None
• Affected tasks: t_079_external_upload_flow
• AI guidelines gap: None