test(corpus): pin v0.3 structural projections

IMPLEMENTATION_PLAN.md

@@ -249,6 +249,15 @@ priorities.
       shared leaf identity. It rejects hostile structural depth before descent
       and leaves redirects, dynamic identities, unproved host command strings,
       and undiscovered executable expressions incomplete.
+- [x] Promote representative existing constructs into the executable corpus
+      with exact v0.3 syntax and command-occurrence expectations. Bash and
+      PowerShell cases pin simple commands, list/pipeline precedence, group
+      scope, static and dynamic wrappers, redirects, hidden substitutions,
+      roles, ancestry, completeness, nullable decoded spans, compatibility
+      operators, and exact shared `Clause` identity. The strict DTO rejects
+      unknown fields and always requires unparseable projections to be empty.
+      The PowerShell manifest now owns all 310 entries and round-trips exactly;
+      explicit false/null assertions remain opt-in and generator-preserved.
 - [ ] Deliver paired Bash and PowerShell `$()` substitution slices for all
       locked executable value positions, including ordering, ancestry,
       shell-specific cwd propagation, literal/escaped boundaries, dynamic
@@ -272,9 +281,6 @@ priorities.
 
 - Seed corpus entries from sanitized real-world dogfood logs (SPEC §14
   workflow) — both shells.
-- Synchronize PowerShell corpus manifest entries 282–309 before using the
-  generator as an all-corpus rewrite; until then, targeted manifest changes
-  must preserve the checked-in hand-authored tail.
 - Expand verb / cmdlet / alias tables as the corpus surfaces real commands.
 - Performance sanity check (~1 ms typical) with a tiny BenchmarkDotNet
   harness — only if anything in the daemon hot path complains.

SPEC.POWERSHELL.md

@@ -1381,6 +1381,13 @@ The shared corpus DTO gains two optional fields:
   `Invoke-Expression`, an over-cap input (§11), or a recursion-depth overflow
   — because real `pwsh` parses the *outer* invocation without error.
 
+The shared v0.3 `syntax` and `commands` expectations defined in `SPEC.md` §13
+apply unchanged. Selected PowerShell entries SHALL pin current-scope groups,
+pipeline/list precedence, exact direct spans, null decoded-wrapper spans,
+command roles, ancestry, completeness, and exact compatibility-leaf identity.
+The manifest marks these entries explicitly so regeneration preserves the
+structural assertions; entries without the marker retain the legacy schema.
+
 ### Coverage targets for v0.2.0
 
 | Category | Min |

SPEC.md

@@ -1964,6 +1964,26 @@ An entry may add an `elements` list to a clause to pin the complete
 `resolved`). The field is opt-in so older corpus entries remain readable;
 issue-specific provenance entries SHALL include it.
 
+An entry may also add both of the following v0.3 structural expectations:
+
+- `syntax` is the complete `ParsedCommand.Syntax` tree flattened in preorder.
+  Each item records `kind`, `parentIndex`, the incoming ancestry `region` and
+  `childIndex`, exact-or-null `sourceStart` / `sourceLength`, and the
+  kind-specific `clauseIndex`, `groupKind`, or `listOperator`. `clauseIndex`
+  identifies the exact compatibility `Clause` instance owned by a
+  `SimpleCommand` node; it is not a copied value comparison.
+- `commands` is the complete `ParsedCommand.Commands` projection in authored
+  order. Each item records its `clauseIndex`, `immediateRole`, `isComplete`,
+  and outermost-to-innermost `ancestry` frames. Each frame records
+  `ancestorKind`, `region`, `childIndex`, and exact-or-null source range.
+
+These fields are independently opt-in so legacy corpus entries retain their
+v0.2 shape; structural acceptance cases normally provide both. When present,
+the runner compares every node, relationship, range, occurrence, role,
+completeness bit, ancestry frame, and `Clause` reference. Unknown JSON members
+are rejected. An unparseable result always asserts empty `Clauses` and
+`Commands`, even when those arrays are omitted from the JSON.
+
 The corpus runner also lexes every direct, parseable input and verifies that
 each authored verb, argument, opaque region, and redirect token is covered by
 exactly-positioned clause-element provenance. This invariant applies even when

TOOLING.md

@@ -49,7 +49,13 @@ and parser slice exists. The PII audit scans both corpus trees.
 
 The curated inputs live in `tools/PwshCorpusTool/CorpusManifest.cs`; the
 `expected` AST is generated from `PwshParser`, and `PwshOracleTests`
-independently validates every input against real `pwsh`.
+independently validates every input against real `pwsh`. The manifest owns
+every checked-in PowerShell entry. `IncludeStructure` opts selected entries
+into exact `syntax` / `commands` expectations, while
+`IncludeOptionalAssertions` preserves explicit false/null security assertions
+for cases that require them. A safe generator change MUST regenerate to a
+temporary directory and produce an exact directory diff before rewriting the
+checked-in corpus.
 
 ## Source Control and CI
 

openspec/changes/v0-3-structured-shell-analysis/specs/executable-command-projection/spec.md

@@ -15,6 +15,12 @@ of whether the command is top-level or nested.
 - **THEN** the authored `echo` command appears once
 - **THEN** its possible effective values are represented by analysis facts rather than three duplicated occurrences
 
+#### Scenario: Executable corpus pins all public projections
+- **WHEN** a selected Bash or PowerShell corpus entry is marked for structural verification
+- **THEN** the corpus records the complete syntax tree and command-occurrence collection in authored order
+- **THEN** every simple-command node and occurrence references the exact compatibility clause by index and object identity
+- **THEN** roles, completeness, ancestry coordinates, and exact-or-null source ranges are compared without weakening legacy corpus entries that omit structural expectations
+
 ### Requirement: Occurrences identify structural execution roles
 Each command occurrence SHALL identify its immediate structural execution role
 and SHALL retain compositional ancestry for analysis, diagnostics, and UI

openspec/changes/v0-3-structured-shell-analysis/tasks.md

@@ -30,7 +30,7 @@
 - [x] 3.6 Adapt the existing Bash grammar to emit the structural model with no newly supported syntax.
 - [x] 3.7 Adapt the existing PowerShell grammar to emit the structural model with no newly supported syntax.
 - [x] 3.8 Add tests proving existing parser inputs retain their v0.2 leaf and compatibility results, except for explicitly promoted v0.3 fail-closed cases.
-- [ ] 3.9 Add corpus expectations for syntax shape, occurrences, roles, and completeness for existing constructs.
+- [x] 3.9 Add corpus expectations for syntax shape, occurrences, roles, and completeness for existing constructs.
 - [ ] 3.10 Implement Bash `$()` discovery in supported argument words, redirect values, iterables, and expanding heredoc bodies; retain literal/escaped spellings and fail closed on command-name substitutions, legacy backticks, or incomplete interiors.
 - [ ] 3.11 Implement PowerShell `$()` discovery in supported words, redirect values, foreach expressions, call-operator dynamic identities, standalone expression statements, double-quoted strings, and expandable here-strings; never invent invocation from standalone output, retain literal/escaped spellings, and fail closed on trailing command-style arguments, call-operator script blocks, or unsupported execution-bearing `@()` / `@{}` forms.
 - [ ] 3.12 Pin substitution parentage, authored sibling indices, innermost-first ordering, Bash-isolated versus PowerShell-current-scope state, unknown-state propagation, nesting/depth limits, and incomplete dynamic identities in direct tests.