0%

Section 23: Full Evaluator Support

Goal: Complete evaluator support for entire Ori spec semantics (parsing assumed working — see Section 0)

SPEC: spec/grammar.ebnf (authoritative), spec/08-types.md, spec/14-expressions.md, spec/09-properties-of-types.md

Status: In Progress — Most features work! 4181 tests pass, 42 skipped. Only a few actual bugs remain.

OVERVIEW

This section ensures the evaluator (interpreter) correctly implements all Ori language semantics. It assumes the parser works correctly (Section 0). The evaluator is in compiler/ori_eval/.

Why this matters: The evaluator is the reference implementation for Ori semantics. It must correctly implement every language feature before LLVM codegen can be validated against it.

Approach:

  1. Audit current evaluator against spec semantics
  2. Implement missing features
  3. Fix incorrect behaviors
  4. Validate with spec tests

23.1 Operators

SPEC: spec/14-expressions.md § Operators

23.1.1 Null Coalesce Operator (??)

Test Status: ALL PASSING — 31/31 tests pass
WEAK TESTS: No #compile_fail negative pins for invalid ?? usage (e.g., int ?? int)
STALE TEST COMMENTS: coalesce.ori header says “26/31 pass” — all 31/31 pass

  • Implement: ?? operator evaluation — 31/31 tests pass

    • Location: ori_eval/src/interpreter/mod.rs — short-circuit logic in eval_binary
    • Semantics: Option<T> ?? T -> T — return inner value if Some, else right operand
    • Semantics: Result<T, E> ?? T -> T — return inner value if Ok, else right operand
    • Short-circuit: Right operand is NOT evaluated if left is Some/Ok
    • Chaining: a ?? b ?? c ?? default works for all None/Some patterns
    • Ori Tests: tests/spec/expressions/coalesce.ori — 31 passed, 0 failed

23.1.2 Comparison Operators for Option/Result

WEAK TESTS: No #compile_fail for comparing incompatible types
STALE TEST COMMENTS: operators_comparison.ori lines 365, 456 say STATUS: Evaluator [BROKEN] for Option ordering and struct equality — both work

  • Implement: <, <=, >, >= for Option types

    • Spec: None < Some(x) for all x — works correctly
    • Verified: 61 tests pass in tests/spec/expressions/operators_comparison.ori
    • Ori Tests: tests/spec/expressions/operators_comparison.ori

23.1.3 Struct Equality with #derive(Eq)

  • Fix: Equality operators for derived structs

    • Verified: #derive(Eq) type Point = { x: int, y: int } with p1 == p2 works
    • Ori Tests: tests/spec/expressions/operators_comparison.ori — 61 tests pass

23.1.4 Shift Overflow Behavior

  • Fix: Left shift overflow panics correctly

    • Spec: 1 << 63 panics due to overflow
    • Verified: assert_panics(f: () -> 1 << 63), assert_panics(f: () -> 1 << 64), assert_panics(f: () -> 1 << -1) all pass
    • Ori Tests: tests/spec/expressions/operators_bitwise.ori — 43 tests pass

23.2 Primitive Trait Methods

SPEC: spec/09-properties-of-types.md § Built-in Traits
STATUS: ALL IMPLEMENTED

Primitives (int, str, bool, float, etc.) implement standard trait methods.

23.2.1 Printable Trait (.to_str())

  • Implement: .to_str() on primitive types

    • int.to_str() — Works: 42.to_str() == "42"
    • str.to_str() — Works
    • bool.to_str() — Works: true.to_str() == "true"
    • float.to_str() — Works
    • Ori Tests: tests/spec/declarations/traits.ori (30 passed), tests/spec/types/existential.ori (8 passed)

23.2.2 Clone Trait (.clone())

  • Implement: .clone() on primitive types

    • int.clone() — Works: let y = x.clone()
    • str.clone() — Works
    • All primitives are cloneable
    • Ori Tests: tests/spec/declarations/traits.ori, tests/spec/types/existential.ori

23.2.3 Hashable Trait (.hash())

  • Implement: .hash() on primitive types

    • int.hash() — Works
    • str.hash() — Works
    • Ori Tests: tests/spec/declarations/traits.ori

23.3 Type Coercion and Indexing

SPEC: spec/14-expressions.md § Index Access
STATUS: Mostly complete — map returns Option, string indexing returns str, non-string keys work
STALE TEST COMMENTS: index_access.ori lines 122-124 say map returns value not Option (fixed); lines 203-204 say returns char not str (fixed)

23.3.1 Map Index Return Type

  • Fix: Map lookup returns Option<V> per spec

    • Verified: map["a"] ?? 0 works; is_none(opt: map["missing"]) works; empty map returns None
    • Ori Tests: tests/spec/expressions/index_access.ori — 34 passed, 1 skipped

23.3.2 Map Non-String Keys

  • Fix: Allow non-string map keys

    • Verified: {1: "one", 2: "two"} works; map[1] ?? "none" returns "one"; int key None detection works
    • Ori Tests: tests/spec/expressions/index_access.ori — test_map_int_key passes

23.3.3 String Index Return Type

  • Fix: String indexing returns str per spec

    • Verified: "hello"[0] returns "h" (str, not char); s[# - 1] returns "o"
    • Ori Tests: tests/spec/expressions/index_access.ori

23.3.4 List Index Assignment

  • Implement: list[i] = value syntax

    • Error: E6080: index assignment (list[i] = x) is not supported; use functional update patterns instead
    • Note: Needs design decision — current error message suggests functional updates
    • Ori Tests: tests/spec/expressions/index_access.ori

  • /tpr-review passed — independent review found no critical or major issues (or all findings triaged)

  • /impl-hygiene-review passed — hygiene review clean. MUST run AFTER /tpr-review is clean.

  • Subsection close-out (23.3) — MANDATORY before starting the next subsection. Run /improve-tooling retrospectively on THIS subsection’s debugging journey (per .claude/skills/improve-tooling/SKILL.md “Per-Subsection Workflow”): which diagnostics/ scripts you ran, where you added dbg!/tracing calls, where output was hard to interpret, where test failures gave unhelpful messages, where you ran the same command sequence repeatedly. Forward-look: what tool/log/diagnostic would shorten the next regression in this code path by 10 minutes? Implement improvements NOW (zero deferral) and commit each via SEPARATE /commit-push using a valid conventional-commit type (build(diagnostics): ... — surfaced by section-23.3 retrospective — build/test/chore/ci/docs are valid; tools(...) is rejected by the lefthook commit-msg hook). Mandatory even when nothing felt painful. If genuinely no gaps, document briefly: “Retrospective 23.3: no tooling gaps”. Update this subsection’s status in section frontmatter to complete.

  • /sync-claude section-close doc sync — verify Claude artifacts across all section commits. Map changed crates to rules files, check CLAUDE.md, canon.md. Fix drift NOW.

  • Repo hygiene check — run diagnostics/repo-hygiene.sh --check and clean any detected temp files.

23.4 Control Flow

SPEC: spec/14-expressions.md § Control Flow
STATUS: Partial — function field calls work; labeled breaks remain

23.4.1 Labeled Break with Value in Nested Loops

  • Implement: Labeled break:name value for nested loops

    • Note: Unlabeled break correctly breaks innermost loop only (not a bug). The original description (“break value not propagating”) was misleading — this is a missing feature (labeled breaks), not broken behavior.
    • Required: loop:name, break:name value, continue:name support in evaluator
    • Ori Tests: tests/spec/expressions/loops.ori — 1 skip: #skip("requires labeled breaks (loop:name, break:name)")

23.4.2 Function Field Calls

DEAD TEST CODE: tests/spec/types/function_types.ori has ALL 355 lines commented out. Function field calls now work; these tests should be revived.

  • Implement: Calling function stored in struct field

    • Syntax: handler.callback(42) where callback: (int) -> str — works correctly
    • Verified: Both ori check and ori run succeed; no crash. Output correct.
    • Ori Tests: tests/spec/types/function_types.ori

  • /tpr-review passed — independent review found no critical or major issues (or all findings triaged)

  • /impl-hygiene-review passed — hygiene review clean. MUST run AFTER /tpr-review is clean.

  • Subsection close-out (23.4) — MANDATORY before starting the next subsection. Run /improve-tooling retrospectively on THIS subsection’s debugging journey (per .claude/skills/improve-tooling/SKILL.md “Per-Subsection Workflow”): which diagnostics/ scripts you ran, where you added dbg!/tracing calls, where output was hard to interpret, where test failures gave unhelpful messages, where you ran the same command sequence repeatedly. Forward-look: what tool/log/diagnostic would shorten the next regression in this code path by 10 minutes? Implement improvements NOW (zero deferral) and commit each via SEPARATE /commit-push using a valid conventional-commit type (build(diagnostics): ... — surfaced by section-23.4 retrospective — build/test/chore/ci/docs are valid; tools(...) is rejected by the lefthook commit-msg hook). Mandatory even when nothing felt painful. If genuinely no gaps, document briefly: “Retrospective 23.4: no tooling gaps”. Update this subsection’s status in section frontmatter to complete.

  • /sync-claude section-close doc sync — verify Claude artifacts across all section commits. Map changed crates to rules files, check CLAUDE.md, canon.md. Fix drift NOW.

  • Repo hygiene check — run diagnostics/repo-hygiene.sh --check and clean any detected temp files.

23.5 Derived Traits

SPEC: spec/10-declarations.md § Attributes
STATUS: ALL IMPLEMENTED
NEEDS TESTS: Only 3 of 7 derivable traits tested (Eq, Clone, Hashable). Missing: Printable, Debug, Default, Comparable.

23.5.1 #derive(Eq) Implementation

  • Fix: Generated equality for structs

    • Compares all fields correctly
    • Works with == and != operators
    • Verified: #derive(Eq) type Point = {...}; assert(eq: p1 == p2) works
    • Ori Tests: tests/spec/expressions/operators_comparison.ori — 61 tests pass

23.5.2 #derive(Clone) Implementation

  • Fix: Generated clone for structs

    • Clones all fields correctly
    • Verified: #derive(Clone) type Point = {...}; let p2 = p1.clone() works
    • Ori Tests: tests/spec/declarations/attributes.ori — 26 tests pass

23.5.3 #derive(Hashable) Implementation

  • Fix: Generated hash for structs

    • Combines hashes of all fields
    • Verified: #derive(Hashable) type Point = {...}; let h = p.hash() works
    • Ori Tests: tests/spec/declarations/attributes.ori

23.6 Stdlib Types and Methods

SPEC: Various stdlib specs

23.6.1 Queue Type

  • Implement: Queue data structure — 6 tests skipped

    • Queue.enqueue(value:)
    • Queue.dequeue() -> Option<T>
    • Queue.peek() -> Option<T>
    • Queue.len() -> int
    • Queue.is_empty() -> bool
    • Queue.clear()
    • Location: library/std/ or evaluator built-ins

23.6.2 Stack Type

  • Implement: Stack data structure — 6 tests skipped

    • Stack.push(value:)
    • Stack.pop() -> Option<T>
    • Stack.peek() -> Option<T>
    • Stack.len() -> int
    • Stack.is_empty() -> bool
    • Stack.clear()
    • Location: library/std/ or evaluator built-ins

23.6.3 String Slice

  • Implement: String slicing — 2 tests skipped

    • str.slice(start:, end:) method
    • str[start..end] syntax
    • Location: Evaluator string operations

23.6.4 Stdlib Utilities

  • Implement: retry/validate — 5 tests skipped

    • retry(attempts:, delay:, op:)
    • validate(value:, rules:)
    • Location: library/std/

23.6.5 Async/Future Support

  • Implement: Future handling — 1 test skipped

    • Async/await or Future handling
    • Location: Evaluator async support

  • /tpr-review passed — independent review found no critical or major issues (or all findings triaged)

  • /impl-hygiene-review passed — hygiene review clean. MUST run AFTER /tpr-review is clean.

  • Subsection close-out (23.6) — MANDATORY before starting the next subsection. Run /improve-tooling retrospectively on THIS subsection’s debugging journey (per .claude/skills/improve-tooling/SKILL.md “Per-Subsection Workflow”): which diagnostics/ scripts you ran, where you added dbg!/tracing calls, where output was hard to interpret, where test failures gave unhelpful messages, where you ran the same command sequence repeatedly. Forward-look: what tool/log/diagnostic would shorten the next regression in this code path by 10 minutes? Implement improvements NOW (zero deferral) and commit each via SEPARATE /commit-push using a valid conventional-commit type (build(diagnostics): ... — surfaced by section-23.6 retrospective — build/test/chore/ci/docs are valid; tools(...) is rejected by the lefthook commit-msg hook). Mandatory even when nothing felt painful. If genuinely no gaps, document briefly: “Retrospective 23.6: no tooling gaps”. Update this subsection’s status in section frontmatter to complete.

  • /sync-claude section-close doc sync — verify Claude artifacts across all section commits. Map changed crates to rules files, check CLAUDE.md, canon.md. Fix drift NOW.

  • Repo hygiene check — run diagnostics/repo-hygiene.sh --check and clean any detected temp files.

23.8 Parser Feature Support (Type Checker/Evaluator)

SPEC: spec/10-declarations.md § Functions, spec/14-expressions.md § Calls
STATUS: Partial — guard clauses work end-to-end; remaining items need type checker + evaluator

These features have working parser support (Section 0.9.1 complete), but need type checker and/or evaluator implementation.

23.8.1 Guard Clauses

Parser Status: Parses correctly (@f (n: int) -> int if n > 0 = n)
Test File: tests/spec/declarations/clause_params.ori

DEAD TEST CODE: tests/spec/declarations/clause_params.ori has ALL tests commented out. Guard clauses now work; these tests should be revived.

  • Type Checker: Verify guard expression returns bool

    • Verified: Guard type checking works correctly

  • Evaluator: Select matching clause based on guard evaluation

    • Verified: @f (n: int) -> int if n > 0 = n * 2; @f (n: int) -> int = 0; dispatches correctly: f(n: 5) returns 10, f(n: -1) returns 0
    • Semantics: Clauses matched top-to-bottom; guard evaluated after pattern match
    • Semantics: If guard is false, try next clause

23.8.2 List Patterns in Function Parameters

Parser Status: Parses correctly (@len ([]: [T]) -> int = 0)
Test File: tests/spec/declarations/clause_params.ori
Verified not working (2026-03-29): E2003: unknown identifier 'tail' — type checker does not extract bindings from list patterns in function parameters

  • Type Checker: Extract bindings from list patterns

    • Location: ori_types/src/infer/ — pattern binding extraction
    • Bindings: [x, ..tail] creates x: T and tail: [T]
    • Empty: [] pattern matches empty list only

  • Evaluator: Destructure list into pattern bindings

    • Location: ori_eval/src/interpreter/ — parameter binding
    • Semantics: Match list structure, bind named elements
    • Failure: If pattern doesn’t match, try next clause

23.8.3 Const Generics

Parser Status: Parses correctly (@f<$N: int>, @f<$N: int = 10>)
Test File: tests/spec/declarations/generics.ori
Verified not working (2026-03-29): $N not recognized in expression context — fails at parse level

  • Type Checker: Make const generic params available in scope

    • Location: ori_types/src/infer/ — generic parameter handling
    • Binding: $N available as compile-time constant in function body
    • Type: Const param has the declared type (int, bool, etc.)

  • Type Checker: Evaluate const generic default values

    • Constraint: Default must be const-evaluable

  • Type Checker: Support const generic constraints in where clauses

    • Syntax: where N > 0, where N > 0 && N <= 100
    • Evaluation: Constraints checked at monomorphization time

  • Evaluator: Substitute const values at call sites

    • Location: ori_eval/src/interpreter/ — generic instantiation

23.8.4 Variadic Parameters

Parser Status: Parses correctly (@sum (nums: ...int))
Test File: tests/spec/declarations/variadic_params.ori (needs creation)
Verified not working (2026-03-29): E2004: expected 1 arguments, found 3 — type checker does not handle variadic expansion

  • Type Checker: Handle variadic parameter types

    • Location: ori_types/src/infer/ — function signature handling
    • Semantics: ...T in parameter position → receives as [T]
    • Constraint: Only one variadic param allowed per function
    • Constraint: Variadic must be last parameter

  • Evaluator: Collect variadic arguments into list

    • Location: ori_eval/src/interpreter/ — call argument handling
    • Semantics: All remaining args collected into [T]
    • Semantics: Zero args → empty list []

23.8.5 Function-Level Contract Enforcement (pre()/post())

Parser Status: Parses correctly (pre(cond | "msg"), post(r -> cond | "msg"))
IR: CheckExpr struct, CheckRange, stored on function definition node
Test File: tests/spec/patterns/run.ori (commented-out tests at lines 140-288)
Verified not working (2026-03-29): pre() is parsed but NOT evaluated at runtime — safe_div(a: 10, b: 0) produces E6001: division by zero instead of contract violation panic

  • Type Checker: Verify pre() condition is bool-typed

    • Location: ori_types/src/infer/expr.rs — function contract handling
    • Semantics: pre(expr) — expr must be bool
    • Semantics: pre(expr | "msg") — msg must be str

  • Type Checker: Verify post() lambda returns bool

    • Semantics: post(r -> expr) — lambda (T) -> bool where T is result type
    • Semantics: post((a, b) -> expr) — tuple destructuring in lambda

  • Evaluator: Execute pre() contracts before function body

    • Location: ori_eval/src/interpreter/ — function call evaluation
    • Semantics: Evaluate condition; if false, panic with message (or default)
    • Semantics: Multiple pre() contracts evaluated in order; first failure panics

  • Evaluator: Execute post() contracts after function body

    • Semantics: Evaluate lambda with result value; if returns false, panic
    • Semantics: Multiple post() contracts evaluated in order; first failure panics

23.8.6 Spread in Function Calls

Parser Status: Parses correctly (sum(...list))
Test File: tests/spec/expressions/function_calls.ori
Verified not working (2026-03-29): E2001: type mismatch: expected 'int', found '[int]' — type checker does not expand spread

  • Type Checker: Verify spread arg matches variadic param type

    • Location: ori_types/src/infer/ — call type checking
    • Constraint: Spread only valid for variadic parameters
    • Constraint: ...expr where expr: [T] spreads into ...T param

  • Evaluator: Expand spread arguments at call site

    • Location: ori_eval/src/interpreter/ — call argument evaluation
    • Semantics: ...list expands to individual elements
    • Semantics: Multiple spreads allowed: fn(...a, ...b)

  • /tpr-review passed — independent review found no critical or major issues (or all findings triaged)

  • /impl-hygiene-review passed — hygiene review clean. MUST run AFTER /tpr-review is clean.

  • Subsection close-out (23.8) — MANDATORY before starting the next subsection. Run /improve-tooling retrospectively on THIS subsection’s debugging journey (per .claude/skills/improve-tooling/SKILL.md “Per-Subsection Workflow”): which diagnostics/ scripts you ran, where you added dbg!/tracing calls, where output was hard to interpret, where test failures gave unhelpful messages, where you ran the same command sequence repeatedly. Forward-look: what tool/log/diagnostic would shorten the next regression in this code path by 10 minutes? Implement improvements NOW (zero deferral) and commit each via SEPARATE /commit-push using a valid conventional-commit type (build(diagnostics): ... — surfaced by section-23.8 retrospective — build/test/chore/ci/docs are valid; tools(...) is rejected by the lefthook commit-msg hook). Mandatory even when nothing felt painful. If genuinely no gaps, document briefly: “Retrospective 23.8: no tooling gaps”. Update this subsection’s status in section frontmatter to complete.

  • /sync-claude section-close doc sync — verify Claude artifacts across all section commits. Map changed crates to rules files, check CLAUDE.md, canon.md. Fix drift NOW.

  • Repo hygiene check — run diagnostics/repo-hygiene.sh --check and clean any detected temp files.

23.7 Section Completion Checklist

STATUS: MOSTLY COMPLETE — 4181 passed, 0 failed, 42 skipped

Exit Criteria: Every Ori spec semantic is correctly implemented in the evaluator. All spec tests must pass — no skipped tests allowed.

Remaining Issues:

  • Labeled break/continue with values in nested loops (23.4.1)
  • List index assignment needs design decision (23.3.4) — currently returns E6080
  • Parser features (23.8) — const generics, variadics, contracts, spread need type checker + evaluator
  • List patterns in function parameters (23.8.2) — type checker binding extraction
  • 4 commented-out test files (~500+ lines) need revival: function_types.ori, clause_params.ori, literals.ori, functions.ori
  • 5 stale status comments in test files need cleanup (coalesce.ori, index_access.ori x2, operators_comparison.ori x2)
  • Missing negative pins (#compile_fail) across all operator test suites

Test Status Comments

Most previously-broken test files now pass. Status comments in some test files are STALE and should be cleaned up.

RESOLVED (all tests pass, STALE STATUS COMMENTS need cleanup):

  • tests/spec/expressions/coalesce.ori — 31/31 pass | STALE: header says “26/31 pass”, lists chaining/map failures
  • tests/spec/expressions/index_access.ori — 34/34 pass | STALE: lines 122-124 say map returns value not Option; lines 203-204 say returns char not str
  • tests/spec/expressions/operators_comparison.ori — 61/61 pass | STALE: line 365 says Option < not implemented; line 456 says struct eq broken
  • tests/spec/expressions/operators_bitwise.ori — 43/43 pass
  • tests/spec/declarations/traits.ori — 30/30 pass
  • tests/spec/types/existential.ori — 8/8 pass
  • tests/spec/expressions/field_access.ori — 30/30 pass

DEAD TEST CODE (commented-out files, features now work — should be revived):

  • tests/spec/types/function_types.ori — ALL 355 lines commented out; function field calls now work
  • tests/spec/declarations/clause_params.ori — ALL tests commented out; guard clauses now work
  • tests/spec/expressions/literals.ori — ALL tests commented out; many literal patterns likely work
  • tests/spec/declarations/functions.ori — ALL tests commented out; basic function tests should work (variadic section won’t)

GENUINELY INCOMPLETE (features not yet implemented):

  • tests/spec/expressions/loops.ori — labeled break value propagation skipped

Notes

  • This section can be worked on in parallel with Section 0 (parser)

  • Evaluator is the reference implementation; LLVM codegen validates against it

  • Some evaluator work overlaps with type checker (Section 1, 2, 3)

  • Stdlib types may be implemented in Ori itself once evaluator is complete

  • All operator evaluations implemented (23.1) — ??, comparisons, equality, shift overflow all work

  • All primitive trait methods registered (23.2) — .to_str(), .clone(), .hash() work

  • Most indexing behaviors correct per spec (23.3) — map returns Option<V>, string indexing returns str, non-string keys work

  • Control flow semantics (23.4) — function field calls work; labeled breaks still missing

  • All derived traits working (23.5) — #derive(Eq, Clone, Hashable) work

  • Stdlib types (23.6) — Queue/Stack not implemented (aspirational, not spec types)

  • Run cargo st tests/ — 4181 passed, 42 skipped (skips are mostly LLVM/capability issues)

  • /tpr-review passed — independent Codex review found no critical or major issues (or all findings triaged)

  • /impl-hygiene-review passed — implementation hygiene review clean (phase boundaries, SSOT, algorithmic DRY, naming). MUST run AFTER /tpr-review is clean.

  • /improve-tooling retrospective completed — MANDATORY at section close, after both reviews are clean. Reflect on the section’s debugging journey (which diagnostics/ scripts you ran, which command sequences you repeated, where you added ad-hoc dbg!/tracing calls, where output was hard to interpret) and identify any tool/log/diagnostic improvement that would have made this section materially easier OR that would help the next section touching this area. Implement every accepted improvement NOW (zero deferral) and commit each via SEPARATE /commit-push. The retrospective is mandatory even when nothing felt painful — that is exactly when blind spots accumulate. See .claude/skills/improve-tooling/SKILL.md “Retrospective Mode” for the full protocol.

  • /tpr-review passed — independent review found no critical or major issues (or all findings triaged)

  • /impl-hygiene-review passed — hygiene review clean. MUST run AFTER /tpr-review is clean.

  • Subsection close-out (23.7) — MANDATORY before starting the next subsection. Run /improve-tooling retrospectively on THIS subsection’s debugging journey (per .claude/skills/improve-tooling/SKILL.md “Per-Subsection Workflow”): which diagnostics/ scripts you ran, where you added dbg!/tracing calls, where output was hard to interpret, where test failures gave unhelpful messages, where you ran the same command sequence repeatedly. Forward-look: what tool/log/diagnostic would shorten the next regression in this code path by 10 minutes? Implement improvements NOW (zero deferral) and commit each via SEPARATE /commit-push using a valid conventional-commit type (build(diagnostics): ... — surfaced by section-23.7 retrospective — build/test/chore/ci/docs are valid; tools(...) is rejected by the lefthook commit-msg hook). Mandatory even when nothing felt painful. If genuinely no gaps, document briefly: “Retrospective 23.7: no tooling gaps”. Update this subsection’s status in section frontmatter to complete.

  • /sync-claude section-close doc sync — verify Claude artifacts across all section commits. Map changed crates to rules files, check CLAUDE.md, canon.md. Fix drift NOW.

  • Repo hygiene check — run diagnostics/repo-hygiene.sh --check and clean any detected temp files.