storage: define session bundle codec contract

[MicroGery]

Summary

Implements PR 1 of the Session Bundle filesystem codec defined in #1528.

This establishes the portable, state-schema-independent contract and deterministic canonicalization primitives that the streaming archive implementation will build on.

What changed

• define the public Session Bundle service, input, result, quota, manifest, digest, and stable error contracts
• define the opaque state-identity descriptor boundary without interpreting Maka state semantics
• encode and decode Manifest V1 as canonical RFC 8785/JCS UTF-8 and reject non-canonical or authority-bearing fields
• define the exact canonical payload tree-record encoding and SHA-256 digest
• support incremental tree hashing for the future streaming reader
• validate raw UTF-8 path ordering, required payload roots, explicit parents, normalized file modes, entry metadata, and aggregate sizes
• fail closed after any rejected incremental entry
• prevent parser diagnostics from exposing attacker-controlled manifest input
• add a dependency boundary test that prevents codec primitives from reaching Maka state schemas
• export the new primitives from the storage package

Tests

• golden fixtures pin the exact canonical Manifest V1 bytes
• golden fixtures pin exact tree-record bytes and the expected SHA-256 digest
• coverage includes empty directories, executable semantics, invalid Unicode, traversal, duplicates, path conflicts, missing parents, invalid metadata, out-of-order streams, payload overflow, and bounded errors
• npm --workspace @maka/storage test
• npm run lint
• npm run format:check
• npm run build
• npm run typecheck

All checks pass on the latest main.

Part of #1528.

[Astro-Han]

Overall, this is a well-scoped implementation of the Session Bundle codec contract and canonicalization primitives. The responsibility boundary is clear, the manifest and canonical-tree golden tests use independent fixtures, and the dependency contract correctly keeps these primitives isolated from Maka state semantics.\n\nOne non-blocking follow-up: validates with , but reconstructs the envelope using a normal property read. An inherited value (for example after prototype pollution) can therefore be serialized as provenance even though it was not an own JSON member. Please gate the returned optional field with the same check and add an inherited-property regression test.\n\nThis is low severity and does not need to block this PR; it should be addressed before untrusted callers are wired into the encoder or as part of the next Session Bundle codec PR.

[Astro-Han]

Correction to my approval review: the inline code identifiers were stripped by shell formatting. decodeEnvelope validates lastCommittedActivationId with Object.hasOwn(...), but reconstructs the envelope using a normal property read. Please gate the returned optional field with the same Object.hasOwn(...) check and add an inherited-property regression test. This remains a non-blocking follow-up.

[MicroGery]

Summary

Implements PR 1 of the Session Bundle filesystem codec defined in #1528.

This establishes the portable, state-schema-independent contract and deterministic canonicalization primitives that the streaming archive implementation will build on.

What changed

• define the public Session Bundle service, input, result, quota, manifest, digest, and stable error contracts
• define the opaque state-identity descriptor boundary without interpreting Maka state semantics
• encode and decode Manifest V1 as canonical RFC 8785/JCS UTF-8 and reject non-canonical or authority-bearing fields
• define the exact canonical payload tree-record encoding and SHA-256 digest
• support incremental tree hashing for the future streaming reader
• validate raw UTF-8 path ordering, required payload roots, explicit parents, normalized file modes, entry metadata, and aggregate sizes
• fail closed after any rejected incremental entry
• prevent parser diagnostics from exposing attacker-controlled manifest input
• add a dependency boundary test that prevents codec primitives from reaching Maka state schemas
• export the new primitives from the storage package

Tests

• golden fixtures pin the exact canonical Manifest V1 bytes
• golden fixtures pin exact tree-record bytes and the expected SHA-256 digest
• coverage includes empty directories, executable semantics, invalid Unicode, traversal, duplicates, path conflicts, missing parents, invalid metadata, out-of-order streams, payload overflow, and bounded errors
• npm --workspace @maka/storage test
• npm run lint
• npm run format:check
• npm run build
• npm run typecheck

All checks pass on the latest main.

Part of #1528.

[Astro-Han]

Overall, this is a well-scoped implementation of the Session Bundle codec contract and canonicalization primitives. The responsibility boundary is clear, the manifest and canonical-tree golden tests use independent fixtures, and the dependency contract correctly keeps these primitives isolated from Maka state semantics.\n\nOne non-blocking follow-up: validates with , but reconstructs the envelope using a normal property read. An inherited value (for example after prototype pollution) can therefore be serialized as provenance even though it was not an own JSON member. Please gate the returned optional field with the same check and add an inherited-property regression test.\n\nThis is low severity and does not need to block this PR; it should be addressed before untrusted callers are wired into the encoder or as part of the next Session Bundle codec PR.

[Astro-Han]

Correction to my approval review: the inline code identifiers were stripped by shell formatting. decodeEnvelope validates lastCommittedActivationId with Object.hasOwn(...), but reconstructs the envelope using a normal property read. Please gate the returned optional field with the same Object.hasOwn(...) check and add an inherited-property regression test. This remains a non-blocking follow-up.

[MicroGery]

Summary

Implements PR 1 of the Session Bundle filesystem codec defined in #1528.

This establishes the portable, state-schema-independent contract and deterministic canonicalization primitives that the streaming archive implementation will build on.

What changed

• define the public Session Bundle service, input, result, quota, manifest, digest, and stable error contracts
• define the opaque state-identity descriptor boundary without interpreting Maka state semantics
• encode and decode Manifest V1 as canonical RFC 8785/JCS UTF-8 and reject non-canonical or authority-bearing fields
• define the exact canonical payload tree-record encoding and SHA-256 digest
• support incremental tree hashing for the future streaming reader
• validate raw UTF-8 path ordering, required payload roots, explicit parents, normalized file modes, entry metadata, and aggregate sizes
• fail closed after any rejected incremental entry
• prevent parser diagnostics from exposing attacker-controlled manifest input
• add a dependency boundary test that prevents codec primitives from reaching Maka state schemas
• export the new primitives from the storage package

Tests

• golden fixtures pin the exact canonical Manifest V1 bytes
• golden fixtures pin exact tree-record bytes and the expected SHA-256 digest
• coverage includes empty directories, executable semantics, invalid Unicode, traversal, duplicates, path conflicts, missing parents, invalid metadata, out-of-order streams, payload overflow, and bounded errors
• npm --workspace @maka/storage test
• npm run lint
• npm run format:check
• npm run build
• npm run typecheck

All checks pass on the latest main.

Part of #1528.

[Astro-Han]

Overall, this is a well-scoped implementation of the Session Bundle codec contract and canonicalization primitives. The responsibility boundary is clear, the manifest and canonical-tree golden tests use independent fixtures, and the dependency contract correctly keeps these primitives isolated from Maka state semantics.\n\nOne non-blocking follow-up: validates with , but reconstructs the envelope using a normal property read. An inherited value (for example after prototype pollution) can therefore be serialized as provenance even though it was not an own JSON member. Please gate the returned optional field with the same check and add an inherited-property regression test.\n\nThis is low severity and does not need to block this PR; it should be addressed before untrusted callers are wired into the encoder or as part of the next Session Bundle codec PR.

[Astro-Han]

Correction to my approval review: the inline code identifiers were stripped by shell formatting. decodeEnvelope validates lastCommittedActivationId with Object.hasOwn(...), but reconstructs the envelope using a normal property read. Please gate the returned optional field with the same Object.hasOwn(...) check and add an inherited-property regression test. This remains a non-blocking follow-up.

[MicroGery]

Summary

Implements PR 1 of the Session Bundle filesystem codec defined in #1528.

This establishes the portable, state-schema-independent contract and deterministic canonicalization primitives that the streaming archive implementation will build on.

What changed

• define the public Session Bundle service, input, result, quota, manifest, digest, and stable error contracts
• define the opaque state-identity descriptor boundary without interpreting Maka state semantics
• encode and decode Manifest V1 as canonical RFC 8785/JCS UTF-8 and reject non-canonical or authority-bearing fields
• define the exact canonical payload tree-record encoding and SHA-256 digest
• support incremental tree hashing for the future streaming reader
• validate raw UTF-8 path ordering, required payload roots, explicit parents, normalized file modes, entry metadata, and aggregate sizes
• fail closed after any rejected incremental entry
• prevent parser diagnostics from exposing attacker-controlled manifest input
• add a dependency boundary test that prevents codec primitives from reaching Maka state schemas
• export the new primitives from the storage package

Tests

• golden fixtures pin the exact canonical Manifest V1 bytes
• golden fixtures pin exact tree-record bytes and the expected SHA-256 digest
• coverage includes empty directories, executable semantics, invalid Unicode, traversal, duplicates, path conflicts, missing parents, invalid metadata, out-of-order streams, payload overflow, and bounded errors
• npm --workspace @maka/storage test
• npm run lint
• npm run format:check
• npm run build
• npm run typecheck

All checks pass on the latest main.

Part of #1528.

[Astro-Han]

Overall, this is a well-scoped implementation of the Session Bundle codec contract and canonicalization primitives. The responsibility boundary is clear, the manifest and canonical-tree golden tests use independent fixtures, and the dependency contract correctly keeps these primitives isolated from Maka state semantics.\n\nOne non-blocking follow-up: validates with , but reconstructs the envelope using a normal property read. An inherited value (for example after prototype pollution) can therefore be serialized as provenance even though it was not an own JSON member. Please gate the returned optional field with the same check and add an inherited-property regression test.\n\nThis is low severity and does not need to block this PR; it should be addressed before untrusted callers are wired into the encoder or as part of the next Session Bundle codec PR.

[Astro-Han]

Correction to my approval review: the inline code identifiers were stripped by shell formatting. decodeEnvelope validates lastCommittedActivationId with Object.hasOwn(...), but reconstructs the envelope using a normal property read. Please gate the returned optional field with the same Object.hasOwn(...) check and add an inherited-property regression test. This remains a non-blocking follow-up.

[MicroGery]

Summary

Implements PR 1 of the Session Bundle filesystem codec defined in #1528.

This establishes the portable, state-schema-independent contract and deterministic canonicalization primitives that the streaming archive implementation will build on.

What changed

• define the public Session Bundle service, input, result, quota, manifest, digest, and stable error contracts
• define the opaque state-identity descriptor boundary without interpreting Maka state semantics
• encode and decode Manifest V1 as canonical RFC 8785/JCS UTF-8 and reject non-canonical or authority-bearing fields
• define the exact canonical payload tree-record encoding and SHA-256 digest
• support incremental tree hashing for the future streaming reader
• validate raw UTF-8 path ordering, required payload roots, explicit parents, normalized file modes, entry metadata, and aggregate sizes
• fail closed after any rejected incremental entry
• prevent parser diagnostics from exposing attacker-controlled manifest input
• add a dependency boundary test that prevents codec primitives from reaching Maka state schemas
• export the new primitives from the storage package

Tests

• golden fixtures pin the exact canonical Manifest V1 bytes
• golden fixtures pin exact tree-record bytes and the expected SHA-256 digest
• coverage includes empty directories, executable semantics, invalid Unicode, traversal, duplicates, path conflicts, missing parents, invalid metadata, out-of-order streams, payload overflow, and bounded errors
• npm --workspace @maka/storage test
• npm run lint
• npm run format:check
• npm run build
• npm run typecheck

All checks pass on the latest main.

Part of #1528.

[Astro-Han]

Overall, this is a well-scoped implementation of the Session Bundle codec contract and canonicalization primitives. The responsibility boundary is clear, the manifest and canonical-tree golden tests use independent fixtures, and the dependency contract correctly keeps these primitives isolated from Maka state semantics.\n\nOne non-blocking follow-up: validates with , but reconstructs the envelope using a normal property read. An inherited value (for example after prototype pollution) can therefore be serialized as provenance even though it was not an own JSON member. Please gate the returned optional field with the same check and add an inherited-property regression test.\n\nThis is low severity and does not need to block this PR; it should be addressed before untrusted callers are wired into the encoder or as part of the next Session Bundle codec PR.

[Astro-Han]

Correction to my approval review: the inline code identifiers were stripped by shell formatting. decodeEnvelope validates lastCommittedActivationId with Object.hasOwn(...), but reconstructs the envelope using a normal property read. Please gate the returned optional field with the same Object.hasOwn(...) check and add an inherited-property regression test. This remains a non-blocking follow-up.

[MicroGery]

Summary

Implements PR 1 of the Session Bundle filesystem codec defined in #1528.

This establishes the portable, state-schema-independent contract and deterministic canonicalization primitives that the streaming archive implementation will build on.

What changed

• define the public Session Bundle service, input, result, quota, manifest, digest, and stable error contracts
• define the opaque state-identity descriptor boundary without interpreting Maka state semantics
• encode and decode Manifest V1 as canonical RFC 8785/JCS UTF-8 and reject non-canonical or authority-bearing fields
• define the exact canonical payload tree-record encoding and SHA-256 digest
• support incremental tree hashing for the future streaming reader
• validate raw UTF-8 path ordering, required payload roots, explicit parents, normalized file modes, entry metadata, and aggregate sizes
• fail closed after any rejected incremental entry
• prevent parser diagnostics from exposing attacker-controlled manifest input
• add a dependency boundary test that prevents codec primitives from reaching Maka state schemas
• export the new primitives from the storage package

Tests

• golden fixtures pin the exact canonical Manifest V1 bytes
• golden fixtures pin exact tree-record bytes and the expected SHA-256 digest
• coverage includes empty directories, executable semantics, invalid Unicode, traversal, duplicates, path conflicts, missing parents, invalid metadata, out-of-order streams, payload overflow, and bounded errors
• npm --workspace @maka/storage test
• npm run lint
• npm run format:check
• npm run build
• npm run typecheck

All checks pass on the latest main.

Part of #1528.

[Astro-Han]

Overall, this is a well-scoped implementation of the Session Bundle codec contract and canonicalization primitives. The responsibility boundary is clear, the manifest and canonical-tree golden tests use independent fixtures, and the dependency contract correctly keeps these primitives isolated from Maka state semantics.\n\nOne non-blocking follow-up: validates with , but reconstructs the envelope using a normal property read. An inherited value (for example after prototype pollution) can therefore be serialized as provenance even though it was not an own JSON member. Please gate the returned optional field with the same check and add an inherited-property regression test.\n\nThis is low severity and does not need to block this PR; it should be addressed before untrusted callers are wired into the encoder or as part of the next Session Bundle codec PR.

[Astro-Han]

Correction to my approval review: the inline code identifiers were stripped by shell formatting. decodeEnvelope validates lastCommittedActivationId with Object.hasOwn(...), but reconstructs the envelope using a normal property read. Please gate the returned optional field with the same Object.hasOwn(...) check and add an inherited-property regression test. This remains a non-blocking follow-up.

[MicroGery]

Summary

Implements PR 1 of the Session Bundle filesystem codec defined in #1528.

This establishes the portable, state-schema-independent contract and deterministic canonicalization primitives that the streaming archive implementation will build on.

What changed

• define the public Session Bundle service, input, result, quota, manifest, digest, and stable error contracts
• define the opaque state-identity descriptor boundary without interpreting Maka state semantics
• encode and decode Manifest V1 as canonical RFC 8785/JCS UTF-8 and reject non-canonical or authority-bearing fields
• define the exact canonical payload tree-record encoding and SHA-256 digest
• support incremental tree hashing for the future streaming reader
• validate raw UTF-8 path ordering, required payload roots, explicit parents, normalized file modes, entry metadata, and aggregate sizes
• fail closed after any rejected incremental entry
• prevent parser diagnostics from exposing attacker-controlled manifest input
• add a dependency boundary test that prevents codec primitives from reaching Maka state schemas
• export the new primitives from the storage package

Tests

• golden fixtures pin the exact canonical Manifest V1 bytes
• golden fixtures pin exact tree-record bytes and the expected SHA-256 digest
• coverage includes empty directories, executable semantics, invalid Unicode, traversal, duplicates, path conflicts, missing parents, invalid metadata, out-of-order streams, payload overflow, and bounded errors
• npm --workspace @maka/storage test
• npm run lint
• npm run format:check
• npm run build
• npm run typecheck

All checks pass on the latest main.

Part of #1528.

[Astro-Han]

Overall, this is a well-scoped implementation of the Session Bundle codec contract and canonicalization primitives. The responsibility boundary is clear, the manifest and canonical-tree golden tests use independent fixtures, and the dependency contract correctly keeps these primitives isolated from Maka state semantics.\n\nOne non-blocking follow-up: validates with , but reconstructs the envelope using a normal property read. An inherited value (for example after prototype pollution) can therefore be serialized as provenance even though it was not an own JSON member. Please gate the returned optional field with the same check and add an inherited-property regression test.\n\nThis is low severity and does not need to block this PR; it should be addressed before untrusted callers are wired into the encoder or as part of the next Session Bundle codec PR.

[Astro-Han]

Correction to my approval review: the inline code identifiers were stripped by shell formatting. decodeEnvelope validates lastCommittedActivationId with Object.hasOwn(...), but reconstructs the envelope using a normal property read. Please gate the returned optional field with the same Object.hasOwn(...) check and add an inherited-property regression test. This remains a non-blocking follow-up.

[MicroGery]

Summary

Implements PR 1 of the Session Bundle filesystem codec defined in #1528.

This establishes the portable, state-schema-independent contract and deterministic canonicalization primitives that the streaming archive implementation will build on.

What changed

• define the public Session Bundle service, input, result, quota, manifest, digest, and stable error contracts
• define the opaque state-identity descriptor boundary without interpreting Maka state semantics
• encode and decode Manifest V1 as canonical RFC 8785/JCS UTF-8 and reject non-canonical or authority-bearing fields
• define the exact canonical payload tree-record encoding and SHA-256 digest
• support incremental tree hashing for the future streaming reader
• validate raw UTF-8 path ordering, required payload roots, explicit parents, normalized file modes, entry metadata, and aggregate sizes
• fail closed after any rejected incremental entry
• prevent parser diagnostics from exposing attacker-controlled manifest input
• add a dependency boundary test that prevents codec primitives from reaching Maka state schemas
• export the new primitives from the storage package

Tests

• golden fixtures pin the exact canonical Manifest V1 bytes
• golden fixtures pin exact tree-record bytes and the expected SHA-256 digest
• coverage includes empty directories, executable semantics, invalid Unicode, traversal, duplicates, path conflicts, missing parents, invalid metadata, out-of-order streams, payload overflow, and bounded errors
• npm --workspace @maka/storage test
• npm run lint
• npm run format:check
• npm run build
• npm run typecheck

All checks pass on the latest main.

Part of #1528.

[Astro-Han]

Overall, this is a well-scoped implementation of the Session Bundle codec contract and canonicalization primitives. The responsibility boundary is clear, the manifest and canonical-tree golden tests use independent fixtures, and the dependency contract correctly keeps these primitives isolated from Maka state semantics.\n\nOne non-blocking follow-up: validates with , but reconstructs the envelope using a normal property read. An inherited value (for example after prototype pollution) can therefore be serialized as provenance even though it was not an own JSON member. Please gate the returned optional field with the same check and add an inherited-property regression test.\n\nThis is low severity and does not need to block this PR; it should be addressed before untrusted callers are wired into the encoder or as part of the next Session Bundle codec PR.

[Astro-Han]

Correction to my approval review: the inline code identifiers were stripped by shell formatting. decodeEnvelope validates lastCommittedActivationId with Object.hasOwn(...), but reconstructs the envelope using a normal property read. Please gate the returned optional field with the same Object.hasOwn(...) check and add an inherited-property regression test. This remains a non-blocking follow-up.