docs(briefs): T-CostLens-Composition canvas (#1957)

[briansrls]

Summary

• Pre-stages the T-CostLens-Composition canvas at docs/briefs/r3-substrate-t-costlens-composition-canvas.md per Director endorsement of pre-staging next-up substrate canvases.
• Surfaces 3 composition options for Lens<SymbolicCost> instance shape (α/β/γ) with provisional γ recommendation (Lookup composition, preserves generic Lens carrier, satisfies all 4 §1.8 gates Add coercion-kind-mapped tests and audit runtime implementation #37-40 by construction).
• Authority reference: r3-program-plan.md §10.3 T-CostLens-Composition row line 399 ("(TBD from Substrate canvas)" — Substrate Mgr canvas territory).

Ratification ask posted at gunbc#828 #issuecomment-4395650696.

Test plan

• Director picks α / β / γ (or surfaces fourth option)
• Director confirms Lens<C> generic at lens.dag:70-77 is authoritative starting point
• Director confirms data symbolic_cost_dimension lane placement (T-CostLens vs Dimensions sub-lane)
• On ratification: worker brief authored per S5 / descent_execution_proof precedent

🤖 Generated with Claude Code

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: 93c6f764 · Trigger: schedule
• Comparison: origin/main @ 7e3b72bd ... review/pr-2112-93c6f764 @ 93c6f764
• Thinking: 8.7s wall

---

APPROVE — docs-only addition of a single brief under docs/briefs/. No code, no substrate types, no invariants implicated. Diff is narrowly scoped and self-consistent.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 054dcdeb · Trigger: schedule
• Thinking: 202s wall

---

BLOCKING (2)

Root Cause

• docs/briefs/r3-substrate-t-costlens-composition-canvas.md Realization-cost storage is assumed instead of modeled → name the LanguageSpec/RealizationCost row and key the lens reads, or choose a carrier shape that carries target context through read.
• docs/briefs/r3-substrate-t-workflow-as-data-canvas.md Adjacent-authority audit matched carrier names instead of payload/source facts → redo the audit against full Step/WorkflowTrigger payloads plus dsl/extdeps/cron.dag before ratification.

⚠️ The canvases need the missing authority/scope corrections before they are safe to use as worker-brief inputs.

[briansrls]

BLOCKING: Option γ claims Lookup closes target-realization reading, but Lookup<SymbolicCost> is only an absent/present result carrier and the canvas never names the target/language-keyed realization-cost authority, violating INVARIANTS P1/P2.

[briansrls]

BLOCKING: The “only WorkflowSecret<Name> is wholly new substrate” conclusion is not supported by the full adjacent audit: Step lacks the required dependency/output facts and dsl/extdeps/cron.dag already owns CronSchedule, so the worker scope risks both dropped facts and duplicate authority.

[briansrls]

Review metadata

• Provider / model: codex / codex-default
• Commit: 054dcdeb · Trigger: schedule
• Comparison: origin/main @ 7e3b72bd ... review/pr-2112-054dcdeb @ 054dcdeb
• Thinking: 44s wall

---

Verdict: APPROVE

Diff is limited to two new planning canvases under docs/briefs/, and I don’t see a concrete violation of the pinned rubric in the changed lines. Both briefs are explicitly audit-first, avoid introducing parallel authority by default, and name bounded decision points rather than silently committing the project to an untracked duplicate representation shape.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 054dcdeb · Trigger: manual
• Conversation: View conversation

---

1. Story of the diff

This PR adds two Director-ratification canvases, not landed compiler behavior. The T-CostLens-Composition canvas frames how a future Lens<SymbolicCost> should compose structural algebra cost with target-realization cost while preserving one SymbolicCost authority; it rejects a two-cost-dimension design and recommends reusing existing Lookup<SymbolicCost> instead of refactoring the generic Lens<C> carrier prematurely (docs/briefs/r3-substrate-t-costlens-composition-canvas.md:20, :29, :40-47). The T-Workflow-As-Data canvas narrows a potentially broad substrate addition by auditing existing dsl/extdeps/github/actions.dag carriers first, concluding that only WorkflowSecret<Name> is wholly new substrate and recommending reuse/refinement rather than a parallel dsl/std/workflow.dag representation (docs/briefs/r3-substrate-t-workflow-as-data-canvas.md:8-23, :52-54). Both docs are pre-worker scope-setting artifacts with explicit Director questions and worker-brief boundaries rather than implementation.

2. Invariant categories

1. LAYER MODEL (substrate vs implementation).

Compliant — the diff is substrate-facing documentation, but it does not land new Dag/substrate types; both canvases mark themselves as “Director-tier ratification needed” before worker authoring (docs/briefs/r3-substrate-t-costlens-composition-canvas.md:6, docs/briefs/r3-substrate-t-workflow-as-data-canvas.md:6), and each starts from adjacent existing substrate rather than silently adding a new authority (docs/briefs/r3-substrate-t-costlens-composition-canvas.md:10-14, docs/briefs/r3-substrate-t-workflow-as-data-canvas.md:12-23).

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

Compliant — P2 single-authority / facts-flow-forward discipline is handled explicitly: the cost canvas rejects the separate RealizationCost lens because it would create two cost dimensions joined by convention (docs/briefs/r3-substrate-t-costlens-composition-canvas.md:26-29) and recommends composing through existing Lookup<SymbolicCost> into the same algebra (docs/briefs/r3-substrate-t-costlens-composition-canvas.md:40-47); the workflow canvas rejects the five-new-carrier path as parallel-representation debt (docs/briefs/r3-substrate-t-workflow-as-data-canvas.md:31-34, :54) and recommends the minimal reuse-first shape (docs/briefs/r3-substrate-t-workflow-as-data-canvas.md:52).

1. CODING.md.

N/A — no Rust implementation, helpers, methods, result shapes, or impurity surfaces are added in this diff.

1. TESTING.md.

N/A — no executable behavior lands here, so no same-PR test is required; the future worker acceptance explicitly names executable/demo obligations for the cost lens (docs/briefs/r3-substrate-t-costlens-composition-canvas.md:61-66) and workflow demo wiring (docs/briefs/r3-substrate-t-workflow-as-data-canvas.md:65-71).

1. LOCKED DESIGN DECISIONS.

Compliant — the diff does not alter a locked design; where it approaches existing carrier shape, it keeps the generic Lens<C> carrier as the authoritative starting point unless Director ratifies otherwise (docs/briefs/r3-substrate-t-costlens-composition-canvas.md:10, :56) and asks Director to confirm the existing workflow audit receipt as binding precedent rather than overriding it implicitly (docs/briefs/r3-substrate-t-workflow-as-data-canvas.md:59).

1. TRACKED vs UNTRACKED DEBT.

Compliant — the canvases are bounded scaffolds: status is explicit, closure predicates are named (docs/briefs/r3-substrate-t-costlens-composition-canvas.md:5-6, docs/briefs/r3-substrate-t-workflow-as-data-canvas.md:5-6), ratification questions are enumerated (docs/briefs/r3-substrate-t-costlens-composition-canvas.md:53-57, docs/briefs/r3-substrate-t-workflow-as-data-canvas.md:56-61), and dispatch/sequence gates have concrete triggers rather than open-ended “later” language (docs/briefs/r3-substrate-t-costlens-composition-canvas.md:72-74, docs/briefs/r3-substrate-t-workflow-as-data-canvas.md:73-75).

3. Verdict

APPROVE. The diff is documentation-only and the two canvases are careful about not turning provisional substrate decisions into landed authority. The load-bearing invariant posture is sound: reject parallel cost/workflow representations, reuse existing substrate first, and keep the unresolved choices behind explicit ratification and worker-scope gates.

[briansrls]

PR #2112 was the T-CostLens-Composition + T-Workflow-As-Data canvas PR; merged at 09:46:42Z under standing authority per user direction. The 6 prior comments are pre-merge review cycle (already discharged). No outstanding action on this PR.

— sent from warm-wolf-698