Section 01: Inventory + owns_crates Normalization

Status: Complete Goal: Produce the factual baseline (current plan landscape + authoritative crate→section map + migration manifest) that §02–§06 depend on. The mechanical routing rule in §02 cannot be written accurately without knowing what Fallback-R looks like in practice; §03’s helper cannot route correctly without complete owns_crates data; §05’s migration cannot execute without a concrete manifest.

Success Criteria:

• Every compiler crate under compiler_repo/compiler/ is enumerated and mapped to exactly one roadmap section’s owns_crates (connects to mission criterion: “Every plans/roadmap/section-*.md has a correct, complete owns_crates: list”)
• Each of the 26 currently-empty owns_crates: [] sections is audited: those that own compiler code get real lists; those that genuinely own no crate keep [] with an inline comment explaining why
• Zero stale crate references remain (including the known ori_lsp phantom in §22-tooling)
• this section’s Manifest Appendix (below) lists every current plans/*/ directory with its classification + target section
• python -m scripts.plan_corpus check plans/roadmap/ exits 0 (returns LOW warning for ori_lsp placeholder directory; documented and accepted as Fallback R carve-out)

Context: The 2026-04-24 bug-tracker redesign established owns_crates: frontmatter as REQUIRED on every roadmap section. The field is universally present (33/33 sections) but 26 sections declare empty lists. /add-bug routes by inverting this frontmatter across all sections — empty lists mean those sections are invisible to routing. Before any plan routing rule lands, the map must be accurate.

Reference implementations:

• Bug-tracker Phase G (2026-04-24): enumerated legacy fix-BUG-*.md sibling files, classified each (completed → archive, in-progress → migrate, orphan → Fallback-R), produced migration manifest before Phase F execution. This section mirrors that sequencing.

Depends on: None (this is Phase 0 prework).

Intelligence Reconnaissance

Queries run 2026-04-24:

• scripts/intel-query.sh --human status — availability probe before running plan-graph queries; if unavailable, fall through to filesystem.
• scripts/intel-query.sh --human plan-status "repr-opt" — pulls section count + blocker count + bug count for the largest in-progress re-route as a smoke-test of the graph’s plan-corpus coverage.
• ls -la plans/ — enumerated 22 re-route directories + roadmap + completed archive. Full inventory captured via Explore agent.
• grep -l "owns_crates:" plans/roadmap/section-*.md | wc -l — 33/33 sections have the field; 26 declare [], 7 declare explicit crate lists.
• grep "absorb-candidate" plans/*/00-overview.md — 8 plans carry migration markers: parser-perf→§00, enum-layout-ssot→§21A, ffi-boundary-safety→§11, repr-opt→§21A, iterator-element-ownership→§07C, locality-representation-unification→§21A, semantic-optimization-pipeline (unmarked), typeck-inference-completeness (§02 implied).

Results summary (≤500 chars) [ori]: 22 re-route plans span a spectrum of scope; 8 carry explicit absorb-candidate markers to specific roadmap sections. owns_crates: coverage is 33/33 present but 26/33 empty; ori_lsp in §22-tooling is stale (crate deleted per project-reorganization). Migration complexity varies: 5 in-progress plans (aot-perf, hygiene-full-2, llvm-verification-tooling, repr-opt, typeck-inference-completeness) need state preservation; 2 research plans (deep-safety, ori-ui-framework) have no section files and need separate handling.

See .claude/skills/query-intel/compose-intel-summary.md for the full query protocol.

01.1 Enumerate current plans/*/ landscape

File(s): this section’s Manifest Appendix (below)

Capture every plans/<dir>/ in machine-readable form so §05 migration can iterate without re-discovery.

• ls -d plans/*/ | grep -v completed → emit list of 22 re-route dirs + roadmap

• For each dir, extract from 00-overview.md frontmatter: status:, plan:, mission paragraph (first paragraph under ## Mission)

• For each dir, count section files: ls plans/<dir>/section-*.md 2>/dev/null | wc -l

• For each dir, check for absorb-candidate marker: grep -h "absorb-candidate" plans/<dir>/*.md || echo "no marker"

• Write inventory to this section’s Manifest Appendix (below) Part A (plan-by-plan table):

  | Plan | Status | Sections | Absorb-candidate | Scope crates |
  |------|--------|----------|------------------|--------------|
  | aot-perf | in-progress | 2 | (none) | ori_llvm, ori_rt |
  | repr-opt | in-progress | 12 | §21A-llvm | ori_canon, ori_arc, ori_repr, ori_llvm, ori_rt |
  | ... (22 rows total) |

• Subsection close-out (01.1) — MANDATORY before starting 01.2:

  • All tasks above are [x] and Manifest Appendix Part A exists and is complete
  • Update this subsection’s status in section frontmatter to complete
  • Repo hygiene check — run compiler_repo/diagnostics/repo-hygiene.sh --check and clean any temp files

01.2 Audit plans/roadmap/section-*.md owns_crates for completeness + staleness

File(s): plans/roadmap/section-*.md (33 files, audit only — no edits yet)

Detect two defect classes: missing real entries, stale entries for deleted crates.

• Enumerate every crate under compiler_repo/compiler/ via ls -d compiler_repo/compiler/*/ → produces authoritative “live crates” set (should include: ori_lexer_core, ori_lexer, ori_parse, ori_ir, ori_types, ori_registry, ori_diagnostic, ori_stack, ori_canon, ori_arc, ori_repr, ori_patterns, ori_eval, ori_llvm, ori_rt, ori_fmt, ori_test_harness, ori_compiler, oric — 19 crates per CLAUDE.md §Compiler)

• For each roadmap section, extract current owns_crates: value

• Compute three sets:

  • Covered: crates appearing in at least one section’s list
  • Uncovered: crates in compiler_repo/compiler/ absent from all sections
  • Stale: entries in any section’s list not matching a live crate (known: ori_lsp in §22)

• Write audit to Manifest Appendix Part B (defect table):

  | Section | Current owns_crates | Defect | Proposed fix |
  |---------|---------------------|--------|--------------|
  | §22-tooling | [ori_fmt, ori_test_harness, ori_stack, ori_lsp, ori_diagnostic, ori_compiler, oric] | STALE: ori_lsp deleted | Remove ori_lsp |
  | §02-type-inference | [] | EMPTY: owns inference code in ori_types | Add ori_types (overlap resolution with §01-type-system in 01.3) |
  | ... |

• Subsection close-out (01.2) — MANDATORY before starting 01.3:

  • Audit complete; Part B of manifest populated
  • Update this subsection’s status to complete
  • Repo hygiene check — compiler_repo/diagnostics/repo-hygiene.sh --check

01.3 Fill the 26 empty owns_crates: [] lists with real crate assignments

File(s): plans/roadmap/section-*.md frontmatter (up to 26 sections edited)

For each section with empty owns_crates: [], decide: does this section own compiler code, and if so, which crates?

Decision criteria:

• Section explicitly about a compiler subsystem (e.g., §02-type-inference) → owns the relevant crate(s)

• Section about a language-feature area that’s stdlib-only (e.g., §07D-stdlib-modules) → may legitimately own [] (stdlib work lives in compiler_repo/library/, not a compiler crate)

• Section about a cross-cutting concern (e.g., §14-testing) → may legitimately own [] (testing crates live under compiler_repo already claimed)

• Resolve crate-overlap: a crate can be owned by exactly ONE section; if §01-type-system currently owns [ori_ir, ori_types] and §02-type-inference’s scope also touches ori_types, decide which section OWNS vs which section references. Edit the LOSING section’s owns_crates to NOT include the disputed crate; add an # Note: comment explaining the overlap resolution

• For each empty section, apply one of: (a) fill with real crate list, (b) retain [] with inline comment # This section owns no compiler crate — work happens in <library/stdlib/crate>, or (c) defer with # TODO-RESOLUTION: crate ownership disputed with §XX, resolved in 01.3 commit

• Remove the stale ori_lsp entry from §22-tooling’s owns_crates

• Verify completeness: every live crate appears in exactly one section’s list; every section’s list references only live crates:

  python3 -c "
  import yaml, pathlib
  live = {p.name for p in pathlib.Path('compiler_repo/compiler').iterdir() if p.is_dir()}
  seen = {}
  for p in sorted(pathlib.Path('plans/roadmap').glob('section-*.md')):
      fm = yaml.safe_load(p.read_text().split('---')[1])
      for c in fm.get('owns_crates', []):
          if c not in live:
              print(f'STALE: {p.name}: {c}')
          if c in seen:
              print(f'DUPLICATE: {c} claimed by {seen[c]} AND {p.name}')
          seen[c] = p.name
  missing = live - set(seen.keys())
  if missing:
      print(f'UNCOVERED: {missing}')
  "

  Expected output: nothing (all stale/duplicate/uncovered defects resolved).

• Commit the normalization as a single logical commit: docs(roadmap): fill empty owns_crates lists + remove stale entries

• Subsection close-out (01.3) — MANDATORY before starting 01.4:

  • Validator script exits 0 (no STALE/DUPLICATE/UNCOVERED output)
  • Normalization commit pushed
  • Update this subsection’s status to complete
  • Repo hygiene check — compiler_repo/diagnostics/repo-hygiene.sh --check

01.4 Produce migration manifest

File(s): this section’s Manifest Appendix (below) (Part C + D)

Using Part A (plan inventory) + the now-complete owns_crates map (output of 01.3), classify every current re-route and decide its target disposition.

Classification algorithm:

1. For plan P with scope crates S = {c1, c2, …}, invert owns_crates map to find owning section per crate
2. Count hits per section: section_hits = {section: |{c in S : c in section.owns_crates}|}
3. If one section has strictly maximum hits → ROUTE A (subsection of that section)
4. If multiple sections tie for max → ROUTE B (Fallback-R with routing_justification)
5. If S ∩ (union of all owns_crates) = ∅ AND S is skill-infra-docs scope → ROUTE C (tracker-index)
6. Special case: research plans with no section files → retained at top level with routing: kind=research-exploration

• For each plan in Part A, apply the classification; write result to Manifest Appendix Part C:

  | Plan | Scope crates | section_hits | Route | Target |
  |------|--------------|--------------|-------|--------|
  | aot-perf | {ori_llvm, ori_rt} | §21A: 1, §21B: 1 (TIE) | B (Fallback-R) | plans/aot-perf/ with routing_justification |
  | repr-opt | {ori_canon, ori_arc, ori_repr, ori_llvm, ori_rt} | §21A: 4, §21B: 1 | A (subsection) | §21A-llvm |
  | parser-perf | {ori_lexer, ori_parse} | §00: 2 | A (subsection) | §00-parser |
  | ... |

• For ROUTE A entries, draft the target subsection ID + title (e.g., §21A.X repr-opt absorption). X is next available subsection number; plan author’s existing section structure becomes sub-subsections.

• For ROUTE B entries, draft the routing_justification text (one sentence explaining why no single section owns this scope).

• For ROUTE C entries (if any), identify target bug-tracker/section-NN-*.md.

• Write Manifest Appendix Part D (state-preservation inventory) for 5 in-progress plans:

  | Plan | review_pipeline | HISTORY blocks | Completed subsections | Blocker cross-refs |
  |------|-----------------|----------------|----------------------|-------------------|
  | aot-perf | rounds_completed: 2, last_round_commit: <sha> | 3 blocks | §01 [x], §02 in-progress | none |
  | repr-opt | (none) | 12 blocks | §01-§05 [x], §06 in-progress | ... |
  | ... |

• Subsection close-out (01.4) — MANDATORY before starting 01.N:

  • Manifest Appendix Parts A–D all populated and reviewed for correctness
  • Update this subsection’s status to complete
  • Repo hygiene check — compiler_repo/diagnostics/repo-hygiene.sh --check

01.N Completion Checklist

• All implementation subsections (01.1–01.4) are [x] and status complete
• All section success criteria have corresponding [x] checkboxes
• Manifest Appendix Parts A–D exist and are populated
• Crate-ownership validator script (from 01.3) exits 0: no STALE, no DUPLICATE, no UNCOVERED
• ./test-all.sh green — regression canary (no compiler code was touched this session — git -C compiler_repo status clean; canary is vacuous by construction)
• python -m scripts.plan_corpus check plans/roadmap/ returns exit 0 (LOW warning for ori_lsp placeholder, non-gating)
• Plan sync — update plan metadata:
  • This section’s frontmatter status → complete, all subsection statuses → complete
  • 00-overview.md Quick Reference table: Section 01 status → Complete
  • 00-overview.md mission success criteria: check off “Every compiler crate is enumerated and mapped” + “26 empty owns_crates lists audited” + “zero stale references”
  • index.md Section 01 status → Complete
• Repo hygiene check — compiler_repo/diagnostics/repo-hygiene.sh --check; clean any temp files

Exit Criteria: Manifest Appendix Parts A–D complete; validator script exits 0; all 26 empty owns_crates audited; ori_lsp removed from §22; ./test-all.sh green; plan_corpus check plans/roadmap/ green.

Manifest Appendix — Parts A/A.1/B/C/D

Machine-readable inventory produced by §01. Consumed verbatim by §05 (migration). Part A/A.1/B/C/D live here; §04 appends Part E, §05 appends Parts F+G, §07 appends Part H — each in their own section’s appendix.

• Part A — plan-by-plan snapshot of plans/*/ (excluding roadmap/, roadmap-plan-routing-unification/, completed/).
• Part A.1 — mission previews (first sentence of ## Mission).
• Part B — owns_crates: defect audit (populated by §01.2).
• Part C — classification + routing decision per plan (populated by §01.4).
• Part D — state-preservation inventory for in-progress plans (populated by §01.4).

Part A — Plan-by-plan inventory

22 re-route plans enumerated 2026-04-24. Absorb-candidate column lists the target plans/roadmap/section-*.md from <!-- phase-g-audit:absorb-candidate --> comments where present (8 plans carry markers). — indicates the plan touches no live compiler crate directly.

Summary:

• Total: 22 re-route plans.
• In-progress: 5 (aot-perf, hygiene-full-2, llvm-verification-tooling, repr-opt, typeck-inference-completeness) — state-preservation required (Part D).
• Research: 2 (deep-safety, ori-ui-framework) — deep-safety has zero section files; ori-ui-framework has 6.
• Absorb-candidate markers: 8 plans carry <!-- phase-g-audit:absorb-candidate --> comments naming a target roadmap section.
• No compiler-crate scope (—): 7 plans (code-journey-rework, deep-safety, ori-ui-framework, pkg_mgmt, project-reorganization, rosetta-stress-test, test-suite-health).

Part A.1 — Mission previews

First sentence of each plan’s ## Mission paragraph (truncated to ≤140 chars).

Part B — owns_crates defect audit

Live-crate set (19, sourced from compiler_repo/compiler/* filtered to directories with Cargo.toml — matches CLAUDE.md §Compiler):

ori_arc, ori_canon, ori_compiler, ori_diagnostic, ori_eval, ori_fmt, ori_ir,
ori_lexer, ori_lexer_core, ori_llvm, ori_parse, ori_patterns, ori_registry,
ori_repr, ori_rt, ori_stack, ori_test_harness, ori_types, oric

Note: compiler_repo/compiler/ori_lsp/ exists as a placeholder directory (.gitkeep only, no Cargo.toml, not a workspace member). It is NOT a live crate.

Coverage analysis (pre-fix):

Non-empty sections (current state):

Defect table:

Fix summary for §01.3:

1. Remove ori_lsp from section-22-tooling.md owns_crates.
2. Add # Note: ... comment to each of the 26 empty sections inline with owns_crates: [], explaining which subsystem section(s) own the relevant crates.
3. The 7 non-empty sections stay as-is (except §22 after step 1).

Part C — Classification + routing decisions

Applied the 6-step classification algorithm from §01.4 to every plan in Part A. section_hits counts how many of a plan’s scope crates are owned by each roadmap section (per post-§01.3 owns_crates:). Strict max → ROUTE A; tie → ROUTE B Fallback-R; no compiler scope → ROUTE C; research with no sections → ROUTE D.

Route summary:

ROUTE A subsection landing table (draft IDs — actual numbers confirmed during §05 migration based on each target section’s next available subsection index):

ROUTE B routing_justification drafts:

ROUTE C skill-infra-docs routing_justification drafts:

Part D — State-preservation inventory for in-progress plans

5 in-progress plans require state preservation before §05 migration (Route A plans are migrated into their owning section; Route B/C/D plans stay peer). State is captured here so the reviewer can verify §05 didn’t silently drop work.

State-preservation protocol for §05 migration:

• ROUTE A migrations: before moving subsections into the owning roadmap section, copy each plan’s 00-overview.md mission + its index.md header into a HISTORY block of the new absorbing subsection. Completed subsections carry their [x] state forward; in-progress ones carry the checkbox state + any review_pipeline: markers verbatim. The old plan directory is moved under plans/completed/ with status: absorbed-by-§XX.Y.
• ROUTE B plans: retain at plans/<plan>/. Add reroute: true + routing_justification: "..." to the plan’s index.md frontmatter per §04/§06 schema. No state change; just a frontmatter marker that /create-plan Step 0.6 recognizes.
• ROUTE C plans: same as ROUTE B — retain + annotate frontmatter with reroute: true.
• ROUTE D plans: retain at top level with routing: kind=research-exploration or equivalent explicit marker. No migration.

Blocker-reference verification checklist for §05 (non-obvious blockers found during Part D scan):

• hygiene-full-2 blocker ref plan/section placeholder — delegated to plans/plan-absorption-followup/section-01-route-a-absorption.md §01.4 (resolution happens during actual absorption; §05 annotation-first scope did not touch blocker refs).
• llvm-verification-tooling references llvm-worker-isolation (Route B) cross-plan dep — delegated to plans/plan-absorption-followup/section-01-route-a-absorption.md §01.6 (verify during absorption).
• repr-opt references bug-tracker/ bug-tracker pointer preservation — delegated to plans/plan-absorption-followup/section-02-repr-opt-flatten.md §02.2 (preserve during umbrella flatten).
• typeck-inference-completeness references BUG-04-091 + BUG-01-002 — delegated to plans/plan-absorption-followup/section-01-route-a-absorption.md per-plan absorption subsection (verify bug entries still exist post-migration).