Fix recoverable attention lifecycle and generic foreground integration

[TKCen]

Summary

• Preserve source-specific attention, bounded leases, held-return checkpoints and fair selection without converting sensing into outbound authority.
• Add typed memory-admission and native-attempt handling, with unsupported automatic memory admission failing closed.
• Expose timestamp-coherent, read-only health projections for attention, execution and sensing.
• Keep deployment identity and reviewer settings in external instance configuration; use generic source, examples, documentation and fixtures.

Verification

The implementation reuses the independently reviewed, sanitized source candidate. Its composed Linux gate passed 1,736 Sensorium tests with one existing held Wiki-integration skip, plus the scoped companion Hermes regressions. The restricted-tool bridge was separately exercised with registered fail-if-called effect sentinels under network isolation; no effect handler ran.

Standalone publication checks on 8a6d08c0703dcb63081bcfb43f63149474c9c905: 1,734 passed, 3 skipped, 0 failures under network isolation; Ruff, wheel and source-distribution builds passed. The skips are two companion-Hermes integration tests and the existing held Wiki integration. The composed suite above separately exercises the companion-Hermes cases. The complete current source tree and both built distributions passed the privacy scan. An existing trailing blank line in a generated CSS file is retained to preserve accepted artifact bytes.

Scope

This PR updates the reusable Sensorium plugin. Private configuration, local installation manifests, audit records and the companion Hermes patch are not published here. No claim of universal prompt-injection resistance or completed long-term behavioural observation is made. Local application and service verification follow the upstream merge.

Current review-repair status

Published review-repair successor ff5f11520ed56353097e81953c5cfa75425891e6.

• Closed the implementation gaps for future-held resume, legacy/malformed lease expiry, cross-root lock ordering, root-confined presentation reads, authored-content retry identity, nonblocking optional observation, frozen settlement labels, and malformed advisory metadata.
• Combined isolated suite: 1,763 passed, 3 explicitly skipped, no failures/errors. Skips require companion host/adapter or held Wiki integration.
• Ruff, package builds, complete current-tree privacy scan, and wheel/sdist member scan passed. Packaged source matches the tested tree.
• The incomplete admission-index prototype is excluded. The lifetime admission scan finding remains unresolved; this successor is not merge-ready and does not imply runtime installation.

The remaining repair must bound physical source reads and ready-state query/projection work without forgetting older disposition authority. Passing tests for bounded file ingestion alone do not close that contract.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-15T12:06:54.363863Z	3a19829	New commits
🔒 Security Review	✅ Completed 2026-09-07T14:01:01.119928Z	a70f190	PR opened

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a70f1906e3

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]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d21fe1d805

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]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8b63f7af67

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]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 34d6802bf6

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".

[TKCen]

@codex review

Please review the final update at 8a6d08c0703dcb63081bcfb43f63149474c9c905. The bounded attention, memory-admission and read-only health changes are now included; standalone tests and package checks pass. Please recheck the prior findings against the updated source.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8a6d08c070

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".

[TKCen]

Published review-repair successor ff5f11520ed56353097e81953c5cfa75425891e6.

• Closed the implementation gaps for future-held resume, legacy/malformed lease expiry, cross-root lock ordering, root-confined presentation reads, authored-content retry identity, nonblocking optional observation, frozen settlement labels, and malformed advisory metadata.
• Combined isolated suite: 1,763 passed, 3 explicitly skipped, no failures/errors. Skips require companion host/adapter or held Wiki integration.
• Ruff, package builds, complete current-tree privacy scan, and wheel/sdist member scan passed. Packaged source matches the tested tree.
• The incomplete admission-index prototype is excluded. The lifetime admission scan finding remains unresolved; this successor is not merge-ready and does not imply runtime installation.

The remaining repair must bound physical source reads and ready-state query/projection work without forgetting older disposition authority. Passing tests for bounded file ingestion alone do not close that contract.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ff5f11520e

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".

[TKCen]

Published the independently accepted bounded-current-attention successor at 9bec616c9e4f8f2848ab8fb7a740adf66a765f07.

• Closed five acceptance cases: receipt-order replay; real query-work bounds; quiet-path full candidate scanning; failed-prepare bypass; and reverse-dependency accuracy with capped fanout.
• Actual SQLite VM-work measurements stayed constant across larger fixtures: prepare 437/437 steps at 10/2,000 archive rows, selection 91/91, disposition ingestion 1,024/1,024 at 10/1,000 current rows, and indexed item lookup 15/15 at 10/5,000 rows.
• Source gate: 1,781 passed with 3 skips (two composition cases and one explicitly excluded Wiki-root case). Final immutable composition separately passed 21/21. These suites overlap; their counts are not additive.
• Independent exact-tree acceptance found no material source findings.

This is source publication evidence only; it does not claim merge, installation, or runtime promotion.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 9bec616c9e

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".

[TKCen]

Published the independently accepted successor at 3a198291a29cbb055f819e2a7afdfb10142ce8b4.

• Closed the remaining input, retry, ownership, outbox and Desktop projection findings. Current attention now remains available as retained history grows, including after provider retry exhaustion, without deleting unresolved items or canonical history.
• Full isolated suite: 1,825 passed, 3 inherited skips. Independent focused review: 112 passed. The composed backend integration separately passed 21/21; these suites overlap and are not additive.
• GitHub CI passed on Python 3.11 and 3.12 for this head. The current-head automated code review is still running.
• The two confirmed defects in the optional default-OFF prospective study are explicitly tracked in Preserve prospective-study observations from short-lived processes #74 and Reject impossible prospective-study cohorts before combinatorial search #75. The study must remain disabled until both are closed; they are deferred, not claimed repaired.

The existing review threads have been reconciled with source/test evidence or the linked study gates. This update does not claim merge, runtime installation, or live lifecycle verification.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3a198291a2

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".