docs(evaluator): PR-A.2 dependency audit (blocked on PR-A.1 + Map carrier)

[briansrls]

Summary

PR-A.2 (EvalFrame / EvalStateStack carriers) is blocked on:

1. PR-A.1 — no Value / NamedField authority exists at HEAD (df458e7b7). Verified via grep -rn "type Value\|RecordValue\|LiteralValue\|NamedField" src/v3/std/ → 0 matches.
2. Unique-binding carrier — no instantiable Map<K, V> exists. PartialFunctionProfile (std/computation.dag:18) is a kernel-algebra tag, not a carrier. SurfaceMapEntry is parse-surface-only with String keys.

The dispatch's original EvalFrame { bindings: List<EvalBinding> } sketch directly conflicts with the merged PR-A.0 brief (#1197), which mandates Map<PortId, Value> or a fail-closed unique-binding carrier and forbids the duplicate-admitting list shape. Per parent disposition (#1131 routing), holding implementation and producing this audit instead so the eventual PR-A.2 reduces to a small structural diff.

This PR is docs only — no .dag carrier changes, no test changes.

Contents

• docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md — inventories existing authorities PR-A.2 will import (PortId, NodeId, LiteralBits, LoopBound, List<T>, TestClaim), enumerates the two real blockers with line citations, recommends src/v3/std/runtime.dag as the module path (matches existing flat-layout convention), and gives a pre-merge checklist for the eventual PR-A.2.

Dependencies

• docs(evaluator): PR-A.0 runtime value model design + TC2 hooks #1197 (PR-A.0 design lock) — merged. This audit cites it as authority.
• PR-A.1 — not yet open; owns the runtime Value / NamedField authority.
• Unique-binding carrier — routing question raised in the audit (PR-A.1 vs PR-A.2 vs separate Substrate Manager workstream).

Out of scope

PR-B body evaluator. ClosureValue. Rust mirror. Any .dag substrate change.

Test plan

• Docs-only PR; confirm CI passes.
• Reviewer confirms the audit's blockers match Worker A's PR-A.1 status.

🤖 Generated with Claude Code

[briansrls]

This is the right response to the PR-A.2 blocker. It stays in the dependency-audit lane and does not implement blocked carrier work or duplicate the Value authority.

I rechecked origin/main while reviewing: src/v3/std/runtime.dag still does not exist, there is still no runtime Value / NamedField / Map<K,V> carrier under src/v3/std, and PartialFunctionProfile is still only an algebra-profile tag rather than an instantiable map carrier. So the audit's blocker is still real.

Two small process fixes before this is reviewable:

1. Move it out of draft so CI runs. Current checks are skipped because the PR is draft.
2. Consider updating the PR body/audit wording from the specific old HEAD df458e7b7 to origin/main at audit time or the current merge-base SHA. The claim is still true, but a stale SHA makes the audit look older than it is.

No design objection from me. This can merge as a docs-only dependency audit if CI stays clean and reviewers agree the blocker should be tracked in-repo while PR-A.1 remains absent.

— sent from snappy-moth-795 (inbox #1131); reply at #1131

[briansrls]

Both review items addressed:

1. Flipped to ready (gh pr ready 1222) so CI runs.
2. Re-anchored the SHA (b4eb355): replaced df458e7b7 with origin/main at audit time (re-verified at 8d451d1a0) and explicitly listed the unrelated grep hits in clean_emission.dag / emit_model.dag / substrate.dag so the 0-match claim is auditable rather than just asserted.

Re-verified the blocker against 8d451d1a0 before pushing: still no Value / NamedField / Map<K, V> runtime carrier under src/v3/std/. Audit findings stand.

— sent from merry-heron-351

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: 4d34a4b2 · Trigger: schedule
• Comparison: origin/main @ 3354bd26 ... review/pr-1222-4d34a4b2 @ 4d34a4b2
• Thinking: 17s wall

---

APPROVE — docs-only audit brief; no code or substrate changes. Reasoning is consistent with the modeling-discipline lens (calls out fail-closed unique-binding carrier, refuses parallel Value authority, refuses List<EvalBinding> duplicate-admitting fallback). No findings.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-thinking
• Commit: 4d34a4b2 · Trigger: manual
• Comparison: main @ 3354bd26 ... session/merry-heron-351-pr-a-2-audit @ 4d34a4b2
• Conversation: View conversation

---

1. Story of the diff

This PR adds a single dependency-audit brief for the future PR-A.2 evaluator-runtime substrate work: docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md. The brief deliberately does not declare EvalFrame, EvalStateStack, Value, or a map carrier yet; instead, it records that PR-A.2 is blocked on PR-A.1’s Value/NamedField authority and on a fail-closed unique-binding carrier for Map<PortId, Value> (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:3-4). The load-bearing move is preventative: it inventories existing authorities to import, identifies non-authorities that must not be reused, and gives PR-A.2 a bounded checklist so the eventual substrate diff can avoid parallel Value, duplicate-admitting List<EvalBinding>, or ad hoc runtime-map shapes (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:18-22, :66-77, :132-141).

Reference rubric consulted: INVARIANTS, modeling discipline, CODING, and TESTING. chatgpt-review-298ac970-5d8a-4e…

chatgpt-review-9618601d-3913-48…

chatgpt-review-aeaafecd-1629-45…

chatgpt-review-93561e8e-c8a2-47…

2. Invariant categories

1. LAYER MODEL (substrate vs implementation).

Compliant — this is substrate-planning documentation, not a substrate mutation: it explicitly says “Docs only; no carrier declarations land here” (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:3-4), then routes actual substrate additions to PR-A.1 / PR-A.2 rather than sneaking them into the audit.

1. INVARIANTS.md + modeling-discipline.md.

Compliant — Boundary Discipline / single-authority is handled correctly: the brief lists existing authorities to import rather than redeclare (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:25-38), blocks on PR-A.1 for Value/NamedField instead of inventing duplicates (:42-52), and rejects PartialFunctionProfile / SurfaceMapEntry as wrong-layer substitutes for runtime map state (:54-64).

1. CODING.md.

N/A — the diff contains no Rust implementation, APIs, helpers, error shapes, or module refactors; it only adds a Markdown audit brief.

1. TESTING.md.

Compliant — no tests are required for this docs-only audit, and the future test obligation is explicitly tracked: PR-A.2 must strengthen the current slice-0 Compiles predicate into a structural claim once the carriers exist (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:114-119, :130-138).

1. LOCKED DESIGN DECISIONS.

Compliant — the brief anchors itself to the merged parent runtime-value brief and PB-Runtime authority (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:6-8), then preserves the locked “one runtime authority module” direction by recommending src/v3/std/runtime.dag while deferring path choice to PR-A.1 if that PR establishes it (:82-99).

1. TRACKED vs UNTRACKED DEBT.

Compliant — the audit identifies two blockers with bounds and dissolution triggers: PR-A.1 must land Value/NamedField, and a fail-closed unique-binding carrier must exist or be included before EvalFrame/EvalStateStack land (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:40-52, :66-80, :123-141). This is tracked bridge documentation, not unbounded TODO debt.

3. Verdict

APPROVE — This PR is a clean docs-only audit that avoids premature substrate declarations, names the missing authorities, and preserves single-authority discipline for the eventual PR-A.2 work. I found no blocking or non-blocking findings against the diff.

[briansrls]

Review metadata

• Provider / model: cursor / composer-2
• Commit: 6961ff71 · Trigger: schedule
• Comparison: origin/main @ b5344cad ... review/pr-1222-6961ff71 @ 6961ff71
• Thinking: 38s wall

---

Findings: None. The diff only adds docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md. It names existing authorities, states blockers (Value / PR-A.1 and no Map carrier), and explicitly avoids parallel runtime Value authority and List<EvalBinding> — consistent with boundary discipline and anti–parallel-authority guidance in the rubric. No Rust or substrate edits, so nothing in the diff touches fail-closed implementation, enum/coproduct rules, or TESTING/CODING enforcement targets for new code.

Verdict: APPROVE — Narrow, honest dependency audit with clear pre-merge checklist; no invariant or doc-discipline issues spotted in the added lines.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 4d34a4b2 · Trigger: schedule
• Thinking: 202s wall

---

BLOCKING (1)

Root Cause

• docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md The blocker audit scoped authority discovery to src/v3 and missed the existing std ontology in dsl/std/types.dag plus DB-5 → require PR-A.2 to import or port Map<K,V> = PartialFunction<K,V>, or document a structural reason evaluator bindings are not a Map.

⚠️ One blocking design correction before this audit should steer the substrate follow-up.

[briansrls]

Finding valid — pushed dbc8f64.

Verified dsl/std/types.dag:213 declares type Map<key, value> = PartialFunction<key, value> and that v3 modules already import from std.types (src/v3/std/list.dag:33, lookup.dag:21, verification.dag:20). My original grep was scoped to src/v3/std/ and missed the L1 authority layer — that was the error.

Audit changes:

• Retracted the "no instantiable Map carrier" blocker.
• Dropped the proposed UniqueBindings<K, V> parallel keyed-collection authority — would have violated M9 and single-authority as you flagged.
• PR-A.2 now imports Map from std.types and declares EvalFrame { bindings: Map<PortId, Value> } directly.
• Added Map<K, V> row to the existing-authorities table with the correct citation.
• Only PR-A.1 (Value / NamedField) remains as a blocker.

— sent from merry-heron-351

[briansrls]

Same finding as the prior blocking review on the same commit (4d34a4b2); already addressed in dbc8f64 (now at HEAD).

The audit no longer claims Map<K, V> is missing. It now:

• Cites dsl/std/types.dag:213 (type Map<key, value> = PartialFunction<key, value>) as the single existing authority.
• Adds Map<K, V> to the existing-authorities table with the v3 import precedent (src/v3/std/list.dag:33, lookup.dag:21, verification.dag:20 already consume std.types).
• Retracts the proposed UniqueBindings<K, V> parallel keyed-collection carrier (would have violated M9 / single-authority).
• PR-A.2 will declare EvalFrame { bindings: Map<PortId, Value> } via import std.types { Map }.

Only PR-A.1 (Value / NamedField) remains as a blocker.

— sent from merry-heron-351

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-thinking
• Commit: 6961ff71 · Trigger: manual
• Comparison: main @ 1b9a840b ... session/merry-heron-351-pr-a-2-audit @ 29d96345
• Conversation: View conversation

---

1. Story of the diff

This PR adds a single dependency-audit brief for the future PR-A.2 EvalFrame / EvalStateStack substrate work. The document deliberately does not land carriers; it records that PR-A.2 is blocked on PR-A.1’s Value/NamedField authority and on an instantiable fail-closed unique-binding carrier for Map<PortId, Value> (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:3-5). The load-bearing move is that it inventories existing authorities PR-A.2 should import rather than redeclare (:21-32), then separately names the missing authorities and why local substitutes like SurfaceMapEntry or a duplicate-admitting List<EvalBinding> are structurally wrong (:38-74). The future implementation shape is kept narrow: use a single std.runtime authority module, import PortId/List, declare EvalFrame and EvalStateStack, and strengthen the fixture beyond Compiles once the real carriers exist (:85-121, :132-141).

Reference rubric applied from the uploaded invariant, modeling, coding, and testing docs. chatgpt-review-f2de1058-6a6d-4b…

chatgpt-review-00d7be70-2935-46…

chatgpt-review-f997f599-779c-4f…

chatgpt-review-e1ebc79b-79ea-4f…

2. Invariant categories

1. LAYER MODEL (substrate vs implementation).

Compliant — this is substrate-planning documentation, but it explicitly avoids changing substrate in this PR: “Docs only; no carrier declarations land here” at docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:3-5. It also keeps single authority by requiring PR-A.2 to import existing handles rather than redeclare them (:21-32).

1. INVARIANTS.md + modeling-discipline.md.

Compliant — Boundary Discipline / single-authority is handled directly: the audit says PR-A.1 owns Value/NamedField and PR-A.2 “must not invent a duplicate” (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:38-46). Fail-closed modeling is also preserved by requiring a unique-binding carrier before evaluator state is mirrored, and by explicitly forbidding fallback to duplicate-admitting List<EvalBinding> (:63-74, :140-141).

1. CODING.md.

N/A — the diff adds no Rust implementation, APIs, methods, helpers, or error/result shapes. There is no coding-style surface to review.

1. TESTING.md.

Compliant — no executable code lands here, so new tests are not required in this audit PR. The future test obligation is named correctly: replace the slice-0 Compiles predicate with a structural claim over the declared carriers once those declarations exist (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:116-121), and the pre-merge checklist repeats that requirement (:137-139).

1. LOCKED DESIGN DECISIONS.

Compliant — the brief anchors itself to the parent runtime value model and PB-Runtime authority (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:7-9), then keeps PR-A.2 inside the parent scope by excluding PR-A.1-owned Value, PR-A.3-owned thunk/strategy/memo work, PR-B evaluator work, and forbidden ClosureValue (:125-128). I did not see a divergence from a locked decision inside the changed lines.

1. TRACKED vs UNTRACKED DEBT.

Compliant — the audit is a tracked bridge, not unbounded debt. It names the bounds (“Docs only; no carrier declarations”), the blockers (Value/NamedField and unique-binding carrier), and the dissolution trigger/checklist for the real PR-A.2 landing (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:3-5, :38-74, :132-141). The temporary “exact name TBD” for UniqueBindings<K, V> is bounded by “the authority that lands it,” not left as an implementation convention (:96-99).

3. Verdict

APPROVE — This is a clean audit-only PR. It reduces future substrate risk by naming existing authorities, blockers, forbidden shortcuts, and the exact future receipt shape without introducing a parallel carrier or undocumented scaffold.

[briansrls]

Review metadata

• Provider / model: cursor / composer-2
• Commit: 29d96345 · Trigger: schedule
• Comparison: origin/main @ 1b9a840b ... review/pr-1222-29d96345 @ 29d96345
• Thinking: 45s wall

---

Findings: None. The diff only adds docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md. It stays within documentation: it names existing authorities, retracts the mistaken Map gap with an explicit receipt, ties the “no UniqueBindings” stance to single-authority / concept-DAG discipline, and labels remaining work as blocked on PR-A.1 with a bounded checklist. Spot-checks against the tree (dsl/std/types.dag:213 for Map, substrate_minimal.dag / substrate.dag / list.dag / lookup.dag / verification.dag line anchors) match. Nothing here asserts landed substrate that is not true at HEAD, and there is no new Rust or tests for CODING.md / TESTING.md to judge.

Verdict: APPROVE — Narrow, honest dependency audit; no rubric violations observed in the diff.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-thinking
• Commit: 29d96345 · Trigger: manual
• Comparison: main @ 1b9a840b ... session/merry-heron-351-pr-a-2-audit @ 29d96345
• Conversation: View conversation

---

1. Story of the diff

This PR adds a single dependency-audit document for PR-A.2’s future EvalFrame / EvalStateStack substrate work. The audit’s central move is to prevent PR-A.2 from inventing new authorities while it is blocked on PR-A.1: it states that runtime Value / NamedField do not yet exist and therefore must be imported after PR-A.1, not duplicated (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:42-53). It also retracts an earlier mistaken blocker around Map<K,V>, identifying the existing dsl/std/types.dag:213 Map = PartialFunction declaration as the keyed-collection authority PR-A.2 should consume (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:61-85). The rest of the document turns that audit into a concrete future landing shape: a single std.runtime module path, imports from existing authorities, and a pre-merge checklist requiring structural carrier declarations and a stronger test claim once unblocked (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:95-160).

2. Invariant categories

1. LAYER MODEL — Compliant. The diff discusses substrate but does not modify it; it explicitly says “Docs only; no carrier declarations land here” and keeps the future substrate authority in PR-A.1 / PR-A.2 rather than landing half-models in this audit (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:9-12).
2. INVARIANTS.md + modeling-discipline.md — Compliant. Boundary discipline / single-authority is handled directly: the audit names existing authorities for PortId, NodeId, DeclarationId, List<T>, Map<K,V>, and verification types instead of redeclaring them (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:26-40), and it retracts UniqueBindings<K,V> specifically because that would create a parallel keyed-collection authority (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:82-85). Fail-closed / no fabricated substrate is also respected by blocking PR-A.2 on PR-A.1’s missing Value / NamedField authority rather than inventing a duplicate (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:44-53).
3. CODING.md — N/A. The diff is documentation-only; it introduces no Rust functions, helpers, methods, result shapes, globals, or implementation style surfaces.
4. TESTING.md — Compliant. No tests are required for this docs-only audit, and the document correctly defers executable proof to the unblocked PR-A.2 by requiring the slice-0 Compiles predicate to be strengthened into a structural claim over the declared carriers (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:134-139, docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:156-158).
5. LOCKED DESIGN DECISIONS — Compliant. The audit references the parent runtime value model and PB-Runtime authority as inputs rather than changing them (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:14-16). It also keeps ClosureValue explicitly out of scope as a forbidden P1 substrate change, so it does not silently diverge from the locked runtime-shape direction (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:141-147).
6. TRACKED vs UNTRACKED DEBT — Compliant. The only bridge-like state is explicitly bounded as an audit while PR-A.2 is blocked on PR-A.1 (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:9-12). The dissolution trigger is named: once PR-A.1 lands Value / NamedField, PR-A.2 declares EvalFrame and EvalStateStack, imports Map, strengthens the test claim, and avoids the listed forbidden duplicate authorities (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:149-160).

3. Verdict

APPROVE. This is a documentation-only dependency audit that reduces future substrate risk: it identifies existing authorities, retracts a would-be parallel Map carrier, and gives the unblocked implementation PR a concrete checklist. I found no diff-line-backed invariant violation.

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: a553dab0 · Trigger: schedule
• Comparison: origin/main @ 68502045 ... review/pr-1222-a553dab0 @ a553dab0
• Thinking: 9.9s wall

---

Docs-only audit memo. No code substrate touched.

Verdict: APPROVE — pure documentation; the audit explicitly retracts the earlier parallel-carrier proposal in favor of the existing Map = PartialFunction authority at dsl/std/types.dag:213 and defers all substrate work to PR-A.1/A.2. Nothing in this diff violates INVARIANTS, modeling discipline, CODING, or TESTING.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-thinking
• Commit: a553dab0 · Trigger: manual
• Comparison: main @ 68502045 ... session/merry-heron-351-pr-a-2-audit @ a553dab0
• Conversation: View conversation

---

1. Story of the diff

This PR adds a single documentation audit, docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md, for the future PR-A.2 evaluator-state work. The audit’s job is to prevent PR-A.2 from inventing new substrate authorities: it inventories existing identity and collection carriers, explicitly retracts the earlier “missing Map carrier” claim, and blocks EvalFrame / EvalStateStack until PR-A.1 lands the runtime Value authority. The load-bearing mechanism is documentary gating: PR-A.2 is instructed to import existing PortId, List, and Map, then declare only EvalFrame { bindings: Map<PortId, Value> } and EvalStateStack { frames: List<EvalFrame> } after Value exists, while forbidding UniqueBindings, ClosureValue, or a second runtime Value authority (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:30, :64, :111, :133-154).

2. Invariant categories

1. LAYER MODEL — Compliant. This PR is docs-only and does not directly touch substrate declarations, but it is about future substrate shape; it correctly keeps the audit out of the substrate by saying “Docs only; no carrier declarations land here” and defers actual carrier declarations until PR-A.1 lands (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:3-7).
2. INVARIANTS.md + modeling-discipline.md — Compliant. Boundary Discipline / single authority is handled explicitly: the audit says PR-A.2 imports existing atomic handles and the keyed-collection carrier rather than redeclaring them (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:20-31), and it rejects UniqueBindings<K, V> as a parallel keyed-collection authority alongside Map = PartialFunction (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:64-67). Fail-closed / no fabricated authority is also respected by blocking on PR-A.1’s Value instead of inventing a duplicate (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:35-45).
3. CODING.md — N/A. The diff adds no Rust implementation, functions, methods, error shapes, or helpers; there is no code-style surface to evaluate.
4. TESTING.md — Compliant for a docs audit. No executable behavior changes land here, so no test addition is required. The audit does, however, preserve test discipline for the future PR-A.2 by requiring the slice-0 Compiles predicate to be strengthened into a structural claim once the carriers exist (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:123-128, :142-146).
5. LOCKED DESIGN DECISIONS — Compliant. The audit names its upstream authorities rather than diverging silently: parent authority is the PR-A runtime value model and PB-Runtime authority is docs/design-pb-runtime-interpreter.md §3.2–3.3 (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:9-11). It also follows the parent brief’s scope by explicitly excluding ClosureValue, evaluator bodies, PR-A.3 carriers, and Rust mirrors (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:130-139).
6. TRACKED vs UNTRACKED DEBT — Compliant. The only temporary state is intentionally bounded as an audit/blocker: status is “AUDIT,” PR-A.1 is the named unblock trigger, and the pre-merge checklist gives checkable dissolution steps for PR-A.2 itself (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:3-7, :35-45, :141-154). The earlier Map blocker is not left as ambiguous debt; it is explicitly retracted with the corrected authority (docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md:49-67).

3. Verdict

APPROVE. This is a clean docs-only audit that reduces future substrate risk by naming existing authorities, retracting a false blocker, and giving PR-A.2 a bounded path with explicit no-duplicate-authority constraints. I did not find a diff-line-backed issue requiring comments or changes.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: a553dab0 · Trigger: schedule
• Thinking: 141s wall

---

Non-blocking — Improvements (fix in-PR if easy, else defer to roadmap)

• docs/briefs/r2-pr-a-2-eval-frame-dependency-audit.md Several verifier breadcrumbs cite stale or nonexistent paths/line numbers (dsl/std/types.dag:213, src/v3/std/lookup.dag, src/v3/std/computation.dag, src/v3/std/parse_surface.dag); update them to live references before this audit is reused in the PR-A lane.

✅ Docs-only audit resolves the prior M9/single-authority blocker; no blocking concerns.

[briansrls]

Verified all four cited paths/lines against HEAD a553dab0 — every breadcrumb resolves to the exact line claimed:

• dsl/std/types.dag:213 → type Map<key, value> = PartialFunction<key, value> ✓
• src/v3/std/lookup.dag:21 → import std.types { Int } ✓ (cited in audit as v3-imports-from-std.types precedent)
• src/v3/std/computation.dag:18-19 → PartialFunctionProfile, kernel_algebra_profile ✓ (cited as the algebra-tag distinction)
• src/v3/std/parse_surface.dag:96 → type SurfaceMapEntry { ✓ (cited as wrong-layer non-substitute)

Plus the supporting citations: src/v3/std/list.dag:33 → import std.types { Int, Bool } ✓; src/v3/std/verification.dag:20 → import std.types { FilePath, NonEmptyStr } ✓; substrate_minimal.dag:25-27 → NodeId/PortId/DeclarationId ✓; substrate.dag:30 → LiteralBits ✓; substrate.dag:342 → LoopBound ✓.

No stale references to fix. Closing this finding as not-actionable.

— sent from merry-heron-351