fix(specs): PR-F-1 supervisor inert-bundle skip + PR-E-1 Cargo cycle (post-#360 corrections)

[AdaWorldAPI]

Summary

Two architectural defects in sprint-3's PR-F-1 and PR-E-1 specs (PR #360 merged) flagged during review but corrections didn't land in the merge window. This PR re-applies the corrections.

Defect 1 — PR-F-1 supervisor inert-bundle handling

The original pre_start loop iterates registry.active_g_list() and unwraps bundle.consumer_pointer. Per the W11 smoke test spec, DOLCE (G=0) and FMA (G=5) are inert bundles with consumer_pointer = None by design. The original loop either panics on unwrap() or aborts pre_start with ActorProcessingErr before any consumer actor spawns.

Fix (append-only correction section at end of pr-f-1-ractor-supervisor.md):

• Explicit match bundle.consumer_pointer.as_ref() { None => continue, Some(p) => spawn } filter
• Two equivalent options documented (Option A inline filter recommended; Option B narrow iterator's contract)
• New test supervisor_skips_inert_bundles_and_spawns_consumers for the W11 smoke-test expectation

Defect 2 — PR-E-1 Cargo dependency cycle

The build script in lance-graph-contract was specified to emit Rust referencing medcare_rs::MedCareActor (and similar for each compiled-in consumer). But every consumer crate depends on lance-graph-contract per Single-Binary Topology I-1 invariant. The build-script reference creates a Cargo dependency cycle (contract → medcare-rs → contract) that won't compile under --features module-medcare.

Fix (append-only correction section at end of pr-e-1-manifest-modules.md):

• Build script in contract emits ONLY data (OGIT::* u32 constants + MANIFEST_METADATA phf::Map of strings/values, no Rust refs to consumer crates)
• Actor registration moves to consumer-side via 3 documented options:
  • Option A (recommended): inventory crate self-registration in each consumer crate
  • Option B: separate umbrella-binary crate that depends on all consumers
  • Option C: callback registry at supervisor init via with_consumers(...)
• Post-fix dependency graph diagram showing no cycles
• Updated PR-E-1 acceptance criteria

What this PR does NOT change

• Pure spec corrections (no code touched)
• Append-only — both correction sections appended at end of respective specs; original PR-F-1 and PR-E-1 content preserved verbatim
• W8 consumer template, W10 sequencing, W11 smoke test, W1 master plan stay valid (corrections internal to PR-F-1 and PR-E-1)

Why a separate PR

PR #360 merged at 21:55 just before my correction pushes landed on the sprint-3 branch (same race as PR #358/#359 sequence in sprint-2). The corrections are append-only, so a follow-up PR is the cleanest path.

Test plan

• PR-F-1 spec has new "CORRECTION (2026-05-12, PR docs(specs): sprint-3 Tier-1 implementation specs (11 PR-X-1 + sequencing + smoke test + consumer template) #360 review)" section with explicit inert-bundle skip
• PR-E-1 spec has new "CORRECTION" section documenting cycle + 3 mechanism options + post-fix dependency graph
• Both correction sections include "Provenance" line crediting user PR docs(specs): sprint-3 Tier-1 implementation specs (11 PR-X-1 + sequencing + smoke test + consumer template) #360 review

Notes

Both defects would have manifested at engineer-pickup time in sprint-4. Catching them at spec stage saves a re-spin loop.

🤖 Generated by Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 38653281e1

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Link inventory submitters into the final binary

With Option A, inventory::submit! records are only available from crates that are actually linked into the executable; making medcare-rs a workspace member or having it depend on lance-graph-callcenter does not cause lance-graph-callcenter/the binary to link medcare-rs. Since this correction removes the contract→consumer edge and explicitly says there is no umbrella-binary requirement, deployments that do not otherwise depend on each consumer crate will see an empty inventory::iter::<ConsumerRegistration>() and spawn no active consumers. The spec needs an explicit binary dependency/static import/features mechanism, or Option B/C should be the recommended path.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Update W8 before marking inventory documented

This marks the W8 consumer-template correction as complete, but .claude/specs/consumer-crate-template.md still omits an inventory dependency in its Cargo.toml example and its actor example only implements Consumer/ConsumerPointer::from_manifest without any inventory::submit! registration. A consumer author following W8 will therefore build a crate that never registers with the new Option A supervisor path, so the acceptance checklist should not be checked until that template is corrected.

Useful? React with 👍 / 👎.