[github-actions]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

@cipherstash/auth@1.0.0

Minor Changes

• fd0ee91: OidcFederationStrategy now keeps one CTS token per distinct provider JWT
  instead of one token for everyone. getJwt is called on every getToken()
  and must return the JWT of the user the current request is for; a JWT is
  exchanged once and its CTS token reused while that token is valid and the
  JWT's entry is still in the bounded least-recently-used cache (eviction before
  expiry, or cacheCapacity: 0, means another exchange; see below). Previously
  the strategy cached the first user's token and handed it to
  every caller until it expired, so on a server serving many users through one
  client a value one user encrypted under a lock context could be bound to
  another user's identity (stack-auth OidcFederationStrategy reuses one user's token for every user until it expires — lock context can bind to the wrong person #1045).

  getToken() now calls getJwt in the caller's own async context and hands the
  JWT to a new getTokenForJwt(jwt) method, on both the Node and wasm-inline
  entries. Previously the Node binding ran getJwt through a napi threadsafe
  function, in the async context of create(), so a callback that read the
  request from AsyncLocalStorage (Clerk's auth(), Next.js headers()) saw
  no request from a module-level strategy, or the creating request from one
  created inside a request. One long-lived strategy now serves every user from
  such a callback. getTokenForJwt is also public, for a caller that already
  holds the user's JWT; it shares the strategy's cache with getToken().

  A token persisted through createWithStore (for example a cookie) is now
  served only to the JWT it was federated from, so a store shared across users,
  or a cookie left over from a previous sign-in, is a cache miss rather than
  another user's token. The store is still one slot, so a store shared by
  several users caches only the most recent of them; give each user their own
  (a per-browser cookie already is). The stored JSON carries a new
  federated_from field; tokens stored by earlier releases have none and are
  re-federated once.

  Because the cache is keyed on the whole JWT, an identity provider that rotates
  JWTs faster than the CTS token lifetime (about 15 minutes) re-federates on each
  rotation, where earlier releases reused the stored token.

  The cache holds 1024 JWTs unless told otherwise. create and createWithStore
  take an optional cacheCapacity after baseUrl (the cacheCapacity option on
  @cipherstash/auth/wasm-inline); when the cache is full the least recently
  used JWT's token is dropped, that user is exchanged again on their next call,
  and the eviction is logged at debug. 0 caches nothing. Existing calls are
  unchanged.

Patch Changes

• db76cf4: @cipherstash/auth's AutoStrategy now honours CS_WORKSPACE_CRN when you are signed in through the CLI (no access key). The default @cipherstash/stack client resolves credentials in @cipherstash/protect-ffi and does not pick this up until that package does. Previously the CRN was ignored and the CLI's current workspace was used, so switching workspaces with the CLI silently changed which workspace a process authenticated against. Now the session saved for the workspace named by the CRN is used, and every token is checked against that workspace, failing with WORKSPACE_MISMATCH otherwise. If no session exists for that workspace, the result is NOT_AUTHENTICATED. Without a CRN, behaviour is unchanged.

  The client key (secretkey.json) that stack-encrypt reads from the CLI profile follows the same workspace, so the token and the client key never come from different ones.

• a1765f1: The native binaries now refuse a TLS certificate chain that carries one of the
  eight eMudhra intermediate CA keys reported compromised on 2026-10-17, including
  their Sectigo cross-signs. On Linux the platform verifier checks no revocation,
  so these intermediates were still trusted through the distro CA bundle. Such a
  chain now fails as a revoked certificate. The package's API is unchanged, and
  the WASM build is unaffected because the runtime's fetch handles TLS there.

• fb75f40: Auth failures carry more help, and their messages never quote a credential or another library's text. A failure's type (NOT_AUTHENTICATED, INVALID_CRN, ...) is unchanged.

  • REQUEST_ERROR's message no longer repeats the transport's own error, which can carry a URL with its query string. It gains help saying what to check.
  • INVALID_TOKEN for a token whose claims do not decode no longer quotes the decoder's message, which could carry a byte or a claim of the token.
  • A failed device binding reports ZeroKMS's status, not its response body.
  • SERVER_ERROR for a refused token exchange names the HTTP status and the auth server's error_description, not the response body, which from the edge in front of it is an HTML page and can echo the access key. A body that is not JSON is reported by where it broke, not by the parser's message.
  • A profile file that is not valid JSON is reported by error kind, line and column, not by the parser's message, which could quote the file.
  • INVALID_GRANT, INVALID_WORKSPACE_ID and ALREADY_CONSUMED gain help, and NOT_AUTHENTICATED's help names stash auth login.
  • A STORE_ERROR carries the help of the profile failure underneath it, such as logging in again when the profile file is missing.

• a076649: Sign-in and token refresh no longer trust an unbounded expires_in from the
  auth server. A token lifetime longer than 30 days is treated as 30 days, and a
  device code is polled for at most one hour, so an oversized value can no
  longer panic the process or wrap to an already-expired token. The buffers that
  carry saved tokens and keys to and from ~/.cipherstash, and an access key
  read from CS_CLIENT_ACCESS_KEY, are now wiped from memory once used.
  A device code's lifetime now runs from when it was issued, and polling stops
  at that deadline rather than at the next poll after it.

• 457b28e: The native binaries now use rustls 0.23.45, which fixes GHSA-2mjx-qc3c-rqvc
  (affecting rustls 0.23.13 up to 0.23.44). The package's API is unchanged.

• b480735: Keep native token refreshes running when their caller is cancelled, so rotated device-session credentials are saved and subsequent requests can still authenticate. Close a refresh-waiter notification race, and give the caller that started a refresh the new token rather than the expiring one it replaced.

  • @cipherstash/auth-darwin-x64@1.0.0
  • @cipherstash/auth-darwin-arm64@1.0.0
  • @cipherstash/auth-win32-x64-msvc@1.0.0
  • @cipherstash/auth-linux-x64-gnu@1.0.0
  • @cipherstash/auth-linux-arm64-gnu@1.0.0
  • @cipherstash/auth-linux-x64-musl@1.0.0

stash@1.3.0

Minor Changes

• bca02b6: stash decrypt status|plan|backfill|drop moves encrypted columns back to plaintext — the encrypt lifecycle run in reverse, with the same safety rails. decrypt plan adds a nullable plaintext twin per EQL v3 column with its type mapped back from the domain (text* → text, timestamp* → timestamptz, json* → jsonb, …): Drizzle schema lines for drizzle-kit generate, or a reviewed SQL migration elsewhere. decrypt backfill decrypts in resumable, checkpointed chunks, one bulkDecrypt (one ZeroKMS request) per chunk, default 500 rows. It refuses a plaintext column that a forward encryption started from and never dropped: only one migration may be open on a pair of columns, so finish or abandon the forward one first. When that column was already removed outside stash, decrypt plan finds it missing and records the forward migration as dropped (marked as observed), so the reverse migration can start. Its first run also refuses a plaintext column that already holds values, because those values may be stale and the backfill only fills NULLs: pass --force to overwrite them from the ciphertext or --confirm-plaintext-current to keep them. Once --force is chosen, or any forced run starts, every later run is forced until a forced pass completes, so an interrupted overwrite is always finished. The application keeps writing both columns until the drop migration is applied. Values bound to an end-user lock context cannot be decrypted with workspace credentials; they are skipped, counted, left NULL and reported, and decrypt drop refuses while any remain. decrypt drop re-verifies coverage at apply time, drops the indexes built on the column's EQL terms, drops the column, and with --uninstall-eql removes EQL v3 when it was the last encrypted column in the database.

  @cipherstash/migrate: runBackfill now runs on a shared chunked engine (runChunkedMigration), which also drives the new runDecryptBackfill. Forward behaviour and recorded event details are unchanged. Adds runDecryptBackfill (resumes after a failed run; PlaintextWriteError keeps decrypted values out of write errors; chunk size capped at MAX_DECRYPT_CHUNK_SIZE), columnHasValues, countUndecrypted, fetchUndecryptedPage, plaintextTypeForEqlDomain, columnSqlType, listColumnIndexes, primaryKeyColumns, countEqlColumnsInDatabase, removeManifestColumn, and the reverse event/phase vocabulary (decrypt_started, decrypt_checkpoint, decrypted, encrypted_dropped; decrypt-dual-writing, decrypting, decrypted, encrypted-dropped), recorded against the encrypted column. MigrationPhase and MigrationEvent gain those members, so an exhaustive switch over them needs the new cases.

Patch Changes

• db76cf4: @cipherstash/auth's AutoStrategy now honours CS_WORKSPACE_CRN when you are signed in through the CLI (no access key). The default @cipherstash/stack client resolves credentials in @cipherstash/protect-ffi and does not pick this up until that package does. Previously the CRN was ignored and the CLI's current workspace was used, so switching workspaces with the CLI silently changed which workspace a process authenticated against. Now the session saved for the workspace named by the CRN is used, and every token is checked against that workspace, failing with WORKSPACE_MISMATCH otherwise. If no session exists for that workspace, the result is NOT_AUTHENTICATED. Without a CRN, behaviour is unchanged.

  The client key (secretkey.json) that stack-encrypt reads from the CLI profile follows the same workspace, so the token and the client key never come from different ones.

• c2798f0: stash encrypt drop and stash decrypt drop now write the Drizzle migration to the directory drizzle-kit reports (the out in drizzle.config.ts) instead of scanning --migrations-dir. Previously, when the two differed, the SQL could be written over an older migration with the same name while drizzle-kit journaled an empty stub in the real out, so the DROP never shipped. When a scan fallback is needed it matches the exact _<name>.sql suffix and refuses to overwrite any file that is not a drizzle-kit scaffold (blank or comment-only), and --migrations-dir now warns when drizzle-kit wrote elsewhere.

• bde5b73: @cipherstash/migrate adds ColumnMigration, one place for the rules of a column migration in either direction: which twin column it pairs with (the flag, the recorded pairing, the manifest, then the naming convention), its phase, whether a forced pass is owed, and which open migration in the other direction blocks it.
  ColumnMigration.load() reads it from cipherstash.cs_migrations; confirmDualWrites(), recordDropGenerated() and recordSourceRemoved() record events, refusing those the rules forbid with a ColumnMigrationError.
  The pure rules are exported as deriveColumnMigration, openCounterpart, planPassFor, passProgress, deriveTwinName and manifestTwin.

  runBackfill and runDecryptBackfill now apply the rules themselves rather than leaving them to the caller.
  Every event records whether its pass was forced.
  An interrupted forced pass is finished as one by the next run, with or without force, and the result says why in forcedBy.
  A forward backfill resumes after a failed run from its last committed position, counts its total once when it resumes a forced pass, and refuses to start while a reverse migration of the same columns is open.
  runBackfill now throws a ColumnMigrationError with code twin-mismatch when encryptedColumn differs from the encrypted column already recorded for that column, rather than continuing into a different one.
  A reverse backfill over a twin that already holds values is forced unless the caller confirmed dual-writes with the plaintext-current waiver.
  Error events store the error's class (errorName) and safe details instead of its message, which could carry row data.

  stash decrypt plan, status, backfill and drop resolve every column through ColumnMigration, so they always agree on the twin and the phase.
  decrypt status flags forward-open, twin-has-values and forced-pass-owed.
  decrypt backfill --confirm-plaintext-current is accepted on a later run as long as no forced pass has started, and refused while a started forced pass is owed.

• b15c1ad: Comparing an encrypted column with a different domain's query operand now raises instead of silently returning no rows. For example public.eql_v3_text_ord = $1::eql_v3.query_text_eq, or eql_v3_text_ord < $1::eql_v3.query_text_ord_ope, previously fell through to native jsonb comparison and answered false for equal values. Within a type family, every pair of a column domain and another domain's query operand now resolves to a blocker for the operators both sides support, naming both operands and telling the caller to use the query domain that matches the column. Use the query domain of the column's own domain (eql_v3.query_<column domain>). The install bundle grows by about 1,060 functions and the same number of operators, so a single-transaction uninstall now takes about 8,550 locks (up from about 6,400); raise max_locks_per_transaction on busy clusters.

• 169f985: Fix: the _ord_ore comparison functions (eql_v3.lt, lte, gt, gte, eq,
  neq), and the operators on _ord_ore columns when named with their schema
  (col OPERATOR(public.<) $1), no longer give wrong answers when public is not
  on the search_path. These functions are kept free of a pinned
  search_path so PostgreSQL can inline them into index scans, which means they
  look up their comparison operator through the caller's search_path. The ORE
  comparison operators live in public, so a caller without it, such as a
  SECURITY DEFINER function with SET search_path = '' (the hardening Supabase
  recommends), got PostgreSQL's built-in record comparison instead. That compares
  the ciphertext bytes, so equality and ordering were silently wrong rather than
  raising an error. Every generated comparison function now names its operator
  with its schema (OPERATOR(public.<) for ORE, OPERATOR(pg_catalog.<) for the
  HMAC, bloom filter and OPE terms), so the result no longer depends on the
  search_path, and the functions still inline. A bare col < $1 still needs
  public on the search_path, because the operators on encrypted columns are
  created in public; without it, call the function (eql_v3.lt(col, $1)) or
  name the operator's schema. Reinstall or upgrade EQL to pick up the fix.

  The stash-postgres skill now explains how to query encrypted columns from a
  function or role whose search_path leaves out public.

• 587d06e: The supply-chain skill names pip among the Dependabot ecosystems the repository monitors, and says that a pip requirements file installed in CI should pin every package with a hash and be installed with pip install --require-hashes --no-deps.

• bd7e55f: Fix invalid YAML in skills/stash-managed-platforms/SKILL.md frontmatter: the
  unquoted description contained a colon-space, so any YAML parser rejected the
  file (this broke cipherstash.com deploys, which mirror these skills at build
  time). Rewritten as a folded block scalar; the parsed description is
  byte-for-byte identical.

• e5a3575: The supply-chain skill names every Dependabot ecosystem the repository monitors, including the Go module and the Docker image the musl binaries are built in.

• 43967d3: Two validation errors no longer quote the caller's search term: a free-text search term that is too short, and a like/ilike pattern with wildcards the encrypted search cannot honour. Each now reports the term's length instead.

  The Supabase and DynamoDB adapters' error logs now record a failure's kind instead of its full message, which could quote a plaintext value. The kind is the error class plus any known encryption error type, FFI error code or auth code it carries; anything else is left out. The DynamoDB options.logger hook now receives a copy of the error whose message is that description, with the same code. The full message is still returned to the caller and passed to errorHandler. @cipherstash/stack/adapter-kit exports the shared describeFailure and likeWildcardError helpers that produce these messages.

• fd0ee91: Document how OidcFederationStrategy serves many users through one client: the
  getJwt callback runs on every operation, in the caller's async context, and
  must return the JWT of the user behind the current request, and each user's
  JWT is exchanged for that user's own token (stack-auth's fix for
  stack-auth OidcFederationStrategy reuses one user's token for every user until it expires — lock context can bind to the wrong person #1045). One caveat is documented for the @cipherstash/stack
  client on the native entry, which calls the strategy from a background task so
  a request read from AsyncLocalStorage is not visible inside getJwt there
  yet: capture the request in the closure or build the client per request until
  @cipherstash/stack runs a second encryption engine (protect-ffi) — new major built on stack-encrypt (umbrella) #1065 lands. The docs and the stash-auth, stash-encryption,
  stash-supabase and stash-edge skills say so, and say that on earlier
  @cipherstash/auth releases a strategy caches one token for everyone, so a
  client there must serve one user.

• ae63dcd: skills/stash-prisma: the stale-migrations/cipherstash/ table now describes
  eql-3.1.0. The baseline bakes 3.1.0, 1.2.x shipped 3.0.6, and the table names
  the db init refusal a stale 1.2.x directory produces.

• bbb1bfa: encrypt on a types.Json column now types its result as Encrypted | null when the plaintext may be null. A null document has always resolved to { data: null } at runtime (it is stored as SQL NULL, not encrypted), but the result was typed Encrypted, so result.data.c compiled and then threw. encrypt is now generic over the plaintext's type: a literal null, or a value typed JsonDocument, gives data: Encrypted | null, through .withLockContext() and .audit() too. A value TypeScript knows is non-null, and every scalar column, still gives data: Encrypted. No runtime change. If your code reads result.data after encrypting a Json value that may be null, it must now check for null first. EncryptOperation and EncryptOperationWithLockContext take an optional result type parameter (default Encrypted), and the EncryptResult<P> helper type is exported from @cipherstash/stack/encryption.

  @cipherstash/stack-supabase: the per-term query-encryption fallback (used when the client has no bulkEncrypt) now rejects a null envelope instead of sending the string "null" as a filter operand, matching the bulk path.

  stash: the stash-encryption skill now says what encrypt returns for a null Json document.

• 8280acf: The stash-dynamodb skill's query examples now narrow the encryptQuery result before reading hm. The previous result.data.hm did not type-check, because the result type also covers ciphertexts, strings and null.

• 9a7815d: The stash-dynamodb skill now scopes the no-re-encryption upgrade to the tested top-level scalar equality columns and unindexed text written by @cipherstash/protect-dynamodb@6.0.1. It explains boolean lookup changes and the privacy cost of a plaintext replacement, stale __hmac removal, downcase normalization, and the required legacy JSON backfill with its rollback limit. It also clarifies legacy v2 reads and queries using current v3 schemas.

• ef3765b: The stash-supabase skill now says that tables with no encrypted columns are
  read and written through the same es.from(), and that an embedded relation to
  an encrypted table is refused rather than decrypted.

• Updated dependencies [db76cf4]

• Updated dependencies [a1765f1]

• Updated dependencies [fb75f40]

• Updated dependencies [a076649]

• Updated dependencies [457b28e]

• Updated dependencies [b480735]

• Updated dependencies [bde5b73]

• Updated dependencies [b15c1ad]

• Updated dependencies [9412408]

• Updated dependencies [37bf4cf]

• Updated dependencies [e930e5b]

• Updated dependencies [0275684]

• Updated dependencies [c851db5]

• Updated dependencies [169f985]

• Updated dependencies [af28159]

• Updated dependencies [fd0ee91]

• Updated dependencies [f7f1c76]

• Updated dependencies [bca02b6]

  • @cipherstash/auth@1.0.0
  • @cipherstash/migrate@1.1.0
  • @cipherstash/eql@3.1.0

@cipherstash/migrate@1.1.0

Minor Changes

• bde5b73: @cipherstash/migrate adds ColumnMigration, one place for the rules of a column migration in either direction: which twin column it pairs with (the flag, the recorded pairing, the manifest, then the naming convention), its phase, whether a forced pass is owed, and which open migration in the other direction blocks it.
  ColumnMigration.load() reads it from cipherstash.cs_migrations; confirmDualWrites(), recordDropGenerated() and recordSourceRemoved() record events, refusing those the rules forbid with a ColumnMigrationError.
  The pure rules are exported as deriveColumnMigration, openCounterpart, planPassFor, passProgress, deriveTwinName and manifestTwin.

  runBackfill and runDecryptBackfill now apply the rules themselves rather than leaving them to the caller.
  Every event records whether its pass was forced.
  An interrupted forced pass is finished as one by the next run, with or without force, and the result says why in forcedBy.
  A forward backfill resumes after a failed run from its last committed position, counts its total once when it resumes a forced pass, and refuses to start while a reverse migration of the same columns is open.
  runBackfill now throws a ColumnMigrationError with code twin-mismatch when encryptedColumn differs from the encrypted column already recorded for that column, rather than continuing into a different one.
  A reverse backfill over a twin that already holds values is forced unless the caller confirmed dual-writes with the plaintext-current waiver.
  Error events store the error's class (errorName) and safe details instead of its message, which could carry row data.

  stash decrypt plan, status, backfill and drop resolve every column through ColumnMigration, so they always agree on the twin and the phase.
  decrypt status flags forward-open, twin-has-values and forced-pass-owed.
  decrypt backfill --confirm-plaintext-current is accepted on a later run as long as no forced pass has started, and refused while a started forced pass is owed.

• bca02b6: stash decrypt status|plan|backfill|drop moves encrypted columns back to plaintext — the encrypt lifecycle run in reverse, with the same safety rails. decrypt plan adds a nullable plaintext twin per EQL v3 column with its type mapped back from the domain (text* → text, timestamp* → timestamptz, json* → jsonb, …): Drizzle schema lines for drizzle-kit generate, or a reviewed SQL migration elsewhere. decrypt backfill decrypts in resumable, checkpointed chunks, one bulkDecrypt (one ZeroKMS request) per chunk, default 500 rows. It refuses a plaintext column that a forward encryption started from and never dropped: only one migration may be open on a pair of columns, so finish or abandon the forward one first. When that column was already removed outside stash, decrypt plan finds it missing and records the forward migration as dropped (marked as observed), so the reverse migration can start. Its first run also refuses a plaintext column that already holds values, because those values may be stale and the backfill only fills NULLs: pass --force to overwrite them from the ciphertext or --confirm-plaintext-current to keep them. Once --force is chosen, or any forced run starts, every later run is forced until a forced pass completes, so an interrupted overwrite is always finished. The application keeps writing both columns until the drop migration is applied. Values bound to an end-user lock context cannot be decrypted with workspace credentials; they are skipped, counted, left NULL and reported, and decrypt drop refuses while any remain. decrypt drop re-verifies coverage at apply time, drops the indexes built on the column's EQL terms, drops the column, and with --uninstall-eql removes EQL v3 when it was the last encrypted column in the database.

  @cipherstash/migrate: runBackfill now runs on a shared chunked engine (runChunkedMigration), which also drives the new runDecryptBackfill. Forward behaviour and recorded event details are unchanged. Adds runDecryptBackfill (resumes after a failed run; PlaintextWriteError keeps decrypted values out of write errors; chunk size capped at MAX_DECRYPT_CHUNK_SIZE), columnHasValues, countUndecrypted, fetchUndecryptedPage, plaintextTypeForEqlDomain, columnSqlType, listColumnIndexes, primaryKeyColumns, countEqlColumnsInDatabase, removeManifestColumn, and the reverse event/phase vocabulary (decrypt_started, decrypt_checkpoint, decrypted, encrypted_dropped; decrypt-dual-writing, decrypting, decrypted, encrypted-dropped), recorded against the encrypted column. MigrationPhase and MigrationEvent gain those members, so an exhaustive switch over them needs the new cases.

Patch Changes

• af28159: The README now calls the forward and reverse moves of a column "column migrations", matching the package glossary, instead of "lifecycle".
• f7f1c76: The package's source links now point at its folder under
  languages/typescript/ in cipherstash/stack. The previous release linked to
  its old folder under packages/, which no longer exists, so the README's
  source links and the npm page's repository link returned 404.

@cipherstash/stack@1.3.0

Minor Changes

• bbb1bfa: encrypt on a types.Json column now types its result as Encrypted | null when the plaintext may be null. A null document has always resolved to { data: null } at runtime (it is stored as SQL NULL, not encrypted), but the result was typed Encrypted, so result.data.c compiled and then threw. encrypt is now generic over the plaintext's type: a literal null, or a value typed JsonDocument, gives data: Encrypted | null, through .withLockContext() and .audit() too. A value TypeScript knows is non-null, and every scalar column, still gives data: Encrypted. No runtime change. If your code reads result.data after encrypting a Json value that may be null, it must now check for null first. EncryptOperation and EncryptOperationWithLockContext take an optional result type parameter (default Encrypted), and the EncryptResult<P> helper type is exported from @cipherstash/stack/encryption.

  @cipherstash/stack-supabase: the per-term query-encryption fallback (used when the client has no bulkEncrypt) now rejects a null envelope instead of sending the string "null" as a filter operand, matching the bulk path.

  stash: the stash-encryption skill now says what encrypt returns for a null Json document.

Patch Changes

• ef3765b: @cipherstash/stack/adapter-kit now exports isEncryptedPayload, the check
  model decryption uses to find an encrypted value in a row. The Supabase adapter
  uses it to refuse a row it would otherwise return undecrypted.

• 43967d3: Two validation errors no longer quote the caller's search term: a free-text search term that is too short, and a like/ilike pattern with wildcards the encrypted search cannot honour. Each now reports the term's length instead.

  The Supabase and DynamoDB adapters' error logs now record a failure's kind instead of its full message, which could quote a plaintext value. The kind is the error class plus any known encryption error type, FFI error code or auth code it carries; anything else is left out. The DynamoDB options.logger hook now receives a copy of the error whose message is that description, with the same code. The full message is still returned to the caller and passed to errorHandler. @cipherstash/stack/adapter-kit exports the shared describeFailure and likeWildcardError helpers that produce these messages.

• fd0ee91: Document how OidcFederationStrategy serves many users through one client: the
  getJwt callback runs on every operation, in the caller's async context, and
  must return the JWT of the user behind the current request, and each user's
  JWT is exchanged for that user's own token (stack-auth's fix for
  stack-auth OidcFederationStrategy reuses one user's token for every user until it expires — lock context can bind to the wrong person #1045). One caveat is documented for the @cipherstash/stack
  client on the native entry, which calls the strategy from a background task so
  a request read from AsyncLocalStorage is not visible inside getJwt there
  yet: capture the request in the closure or build the client per request until
  @cipherstash/stack runs a second encryption engine (protect-ffi) — new major built on stack-encrypt (umbrella) #1065 lands. The docs and the stash-auth, stash-encryption,
  stash-supabase and stash-edge skills say so, and say that on earlier
  @cipherstash/auth releases a strategy caches one token for everyone, so a
  client there must serve one user.

• 4b14ee6: Fix the README's OIDC federation example, which passed the Result from OidcFederationStrategy.create() straight to config.authStrategy and failed in Encryption() with opts.strategy.getToken is missing.
  It now unwraps the Result and passes strategy.data, and so does the LockContext JSDoc example.

• 20ccab3: Internal type cleanup of the encryption client (no runtime change). The exported operation classes EncryptOperation, EncryptQueryOperation, BatchEncryptQueryOperation and BulkEncryptOperation now declare that they accept a types.Json document as plaintext, which they always did at runtime; their getOperation() return types widen to match. The EncryptionClient interface changes only in what encrypt returns for a types.Json column; see the EncryptResult changeset.

• Updated dependencies [db76cf4]

• Updated dependencies [a1765f1]

• Updated dependencies [fb75f40]

• Updated dependencies [a076649]

• Updated dependencies [457b28e]

• Updated dependencies [b480735]

• Updated dependencies [fd0ee91]

• Updated dependencies [457b28e]

• Updated dependencies [f7f1c76]

  • @cipherstash/auth@1.0.0
  • @cipherstash/protect-ffi@0.33.1

@cipherstash/stack-prisma@1.3.0

Minor Changes

• ae63dcd: Move the bundled EQL v3 migrations to eql-3.1.0. Its install SQL changes
  the version stamp (eql_v3.version() returns '3.1.0' and the eql_v3
  schema comment reads '3.1.0') and the ORE block comparator behind every
  ordered ORE operator and the ORE operator class. The comparator now runs a
  constant number of operations whatever two values share, closing a timing
  leak. Ordering results are unchanged for well-formed values, but a block-ORE
  text value with a malformed or mismatched-width term now raises on every
  comparison against it. Check your _ore text columns before migrating; see
  U-001.
  No domain, wire format or index term changes.

  Two artefacts carry the new bundle:

  • A new upgrade edge, 20261005T0000_upgrade_eql_v3_3_1_0, carrying the
    invariant cipherstash:upgrade-eql-v3-bundle-3.1.0-v1. Existing databases
    re-install through it on the next prisma-next migration plan followed by
    prisma-next migrate. migrate alone is not enough: only
    migration plan copies the new directory into your repo.
  • The baseline install migration 20260601T0100_install_eql_v3_bundle, which
    now bakes 3.1.0 and gains a sixth op, a no-SQL carrier for the new
    invariant, so fresh databases land on 3.1.0 from the single all-additive
    genesis edge and db init keeps working.

  A 1.2.x database walks one edge. A 1.0.0 or 1.1.x database walks three.
  1.2.x shipped eql-3.0.6, so migrate walks only the 3.1.0 edge. 1.0.0 and
  1.1.x shipped eql-3.0.4, so migrate walks the 3.0.5, 3.0.6 and 3.1.0 edges in
  one transaction, re-installing the ~2.6 MB bundle three times. As with every EQL
  upgrade, each re-install opens with DROP SCHEMA IF EXISTS eql_v3 CASCADE:
  re-run your grant script, and recreate functional indexes and anything else
  depending on eql_v3.

  Action required, including on 1.2.x: the baseline changed again. 1.2.0
  and 1.2.1 shipped the baseline at eql-3.0.6, and this release changes those
  published bytes. 1.0.0 to 1.1.1 shipped it at eql-3.0.4. If your project has a
  migrations/cipherstash/ directory generated against any earlier version,
  delete it and re-run prisma-next migration plan:

  rm -rf migrations/cipherstash
  npx prisma-next migration plan

  Your database keeps its markers, so already-applied invariants are not re-run.
  If you skip the delete, a fresh db init from a 1.2.x directory refuses with
  Operation cipherstash.upgrade-eql-v3-bundle-3.1.0 has class "data" which is not allowed by policy. From a 1.0.0 or 1.1.x directory it refuses on the 3.0.5
  edge instead. See "Upgrading from 1.0.0, 1.1.x or 1.2.x" in the package README.

  Why the baseline was re-emitted rather than given a second genesis edge.
  The 3.0.5 and 3.0.6 entries said this trade must be re-argued on adoption
  numbers at each bump. Unlike the 3.0.6 re-emit, this one changes a published
  baseline. The team chose it on these numbers: @cipherstash/stack-prisma had
  481 npm downloads in the 30 days to 3 October 2026, and 384 in the last 7 of
  those days, 305 of them on the 1.2.0 release day. The append-only alternative,
  a second from: null genesis edge, would add another permanent ~2.6 MB copy of
  the bundle. That choice was made when 3.1.0 changed only the version stamp; the
  comparator change came after it and does not alter the trade, since every
  database reaches the same bundle either way.

Patch Changes

• 43967d3: Two validation errors no longer quote the caller's search term: a free-text search term that is too short, and a like/ilike pattern with wildcards the encrypted search cannot honour. Each now reports the term's length instead.

  The Supabase and DynamoDB adapters' error logs now record a failure's kind instead of its full message, which could quote a plaintext value. The kind is the error class plus any known encryption error type, FFI error code or auth code it carries; anything else is left out. The DynamoDB options.logger hook now receives a copy of the error whose message is that description, with the same code. The full message is still returned to the caller and passed to errorHandler. @cipherstash/stack/adapter-kit exports the shared describeFailure and likeWildcardError helpers that produce these messages.

• Updated dependencies [ef3765b]

• Updated dependencies [b15c1ad]

• Updated dependencies [9412408]

• Updated dependencies [37bf4cf]

• Updated dependencies [e930e5b]

• Updated dependencies [0275684]

• Updated dependencies [c851db5]

• Updated dependencies [169f985]

• Updated dependencies [43967d3]

• Updated dependencies [fd0ee91]

• Updated dependencies [4b14ee6]

• Updated dependencies [bbb1bfa]

• Updated dependencies [20ccab3]

  • @cipherstash/stack@1.3.0
  • @cipherstash/eql@3.1.0

@cipherstash/eql@3.1.0

Minor Changes

• e930e5b: Security: the ORE block comparator (eql_v3_internal.compare_ore_block_256_term
  and compare_ore_block_256_terms, behind every ordered ORE operator and the ORE
  operator class) now runs a constant number of operations for operands of the
  same shape (term count, term width, NULL positions), whether and wherever they
  differ. Previously its run time depended on where two values first differ, so a
  party able to time queries could learn how long a prefix their query value
  shares with stored values. For text the effect was large: a comparison took
  about 11 µs when the first term differs and 47 µs when only the sixth does.
  Results are unchanged. Single-term comparisons cost about 6% more when the
  terms differ and about 36% more (roughly 1 µs) when they are equal, since equal
  terms now run the hash too; text comparisons now take about 18 µs whatever the
  operands. Multi-term (text) values now check every paired term, so a
  malformed or mismatched-width term after the first difference, or one paired
  with a NULL element, raises instead of being skipped; see
  U-001
  for how to find such rows before upgrading. A timing harness is in
  packages/eql/tests/timing/ore_block_256/.
• 0275684: The eql-bindings crate resolves an EQL type named as a string to its own Stack Encrypt plan (stack-encrypt feature). eql_bindings::encryption::targets carries a catalog-generated table of every EQL type a data plan may name as a field target — its name across languages, family and suffix, the plaintext ValueKind it takes, the indexes it carries, its query twin, and whether the engine can produce it today with the reason when not — and encrypt / decrypt / query entry points that dispatch on the name and run the type's derived EncryptFrom / DecryptInto, resolving to the EQL value's JSON bytes through the engine's Pending so a guest batches it with the rest of a plan. This is the EQL half of EQL types as plan field targets (Go and Rust plans cannot target EQL domains, so Go writes rows EQL cannot query — make domains plan field targets #1062): a Go data plan names TextEq and the guest returns the finished EQL value instead of assembling one. TextEq is the only producible type; every other name is refused with the plan's reason. The SQL surface and the TypeScript package are unchanged.
• c851db5: Rust TextEq and TextEqQuery support Stack Encrypt's target-directed API behind the stack-encrypt feature. The stored table and column identifier supplies the encryption context; native ciphertext and equality terms are transcoded into EQL payloads while Vitamin C owns plaintext encoding. This is a new producer profile with exact string equality, independent of existing cipherstash-client ciphertext and terms. Query operands carry no recoverable ciphertext.

Patch Changes

• b15c1ad: Comparing an encrypted column with a different domain's query operand now raises instead of silently returning no rows. For example public.eql_v3_text_ord = $1::eql_v3.query_text_eq, or eql_v3_text_ord < $1::eql_v3.query_text_ord_ope, previously fell through to native jsonb comparison and answered false for equal values. Within a type family, every pair of a column domain and another domain's query operand now resolves to a blocker for the operators both sides support, naming both operands and telling the caller to use the query domain that matches the column. Use the query domain of the column's own domain (eql_v3.query_<column domain>). The install bundle grows by about 1,060 functions and the same number of operators, so a single-transaction uninstall now takes about 8,550 locks (up from about 6,400); raise max_locks_per_transaction on busy clusters.

• 9412408: Drop the CREATE CAST (public.eql_v3_json_search AS eql_v3.query_json) that PostgreSQL always ignored (it warned "cast will be ignored because the source data type is a domain" on every install). doc::eql_v3.query_json was never converting; call eql_v3.to_ste_vec_query(col) instead. Also corrects the eql-bindings wire-format docs: op backs _ord / _ord_ope (and text_ord / text_search carry hm + op), ob only _ord_ore.

• 37bf4cf: Encrypted JSON arrays now agree between the Rust/TypeScript bindings and the SQL. SteVecDocument models the root a array flag that the SQL root validator accepts and the array accessors require, so an array document the database accepts also parses in the bindings. The a flag on documents and entries is now optional but not nullable in the generated TypeScript and JSON Schema (a?: boolean), matching the SQL CHECK; a null is rejected instead of failing later with a cast error. eql_v3.jsonb_array_length returns 0 for an empty array (it returned NULL), and eql_v3.jsonb_array_elements and eql_v3.jsonb_document_contains(jsonb[], jsonb) no longer raise "upper bound of FOR loop cannot be null" on one.

• 169f985: Fix: the _ord_ore comparison functions (eql_v3.lt, lte, gt, gte, eq,
  neq), and the operators on _ord_ore columns when named with their schema
  (col OPERATOR(public.<) $1), no longer give wrong answers when public is not
  on the search_path. These functions are kept free of a pinned
  search_path so PostgreSQL can inline them into index scans, which means they
  look up their comparison operator through the caller's search_path. The ORE
  comparison operators live in public, so a caller without it, such as a
  SECURITY DEFINER function with SET search_path = '' (the hardening Supabase
  recommends), got PostgreSQL's built-in record comparison instead. That compares
  the ciphertext bytes, so equality and ordering were silently wrong rather than
  raising an error. Every generated comparison function now names its operator
  with its schema (OPERATOR(public.<) for ORE, OPERATOR(pg_catalog.<) for the
  HMAC, bloom filter and OPE terms), so the result no longer depends on the
  search_path, and the functions still inline. A bare col < $1 still needs
  public on the search_path, because the operators on encrypted columns are
  created in public; without it, call the function (eql_v3.lt(col, $1)) or
  name the operator's schema. Reinstall or upgrade EQL to pick up the fix.

  The stash-postgres skill now explains how to query encrypted columns from a
  function or role whose search_path leaves out public.

@cipherstash/nextjs@4.4.1

Patch Changes

• 3bd3d28: Harden the CipherStash session cookie (__cipherstash_cts_session).

  • The cookie is now written httpOnly and secure. It holds a bearer token and nothing in this package reads it client-side; if your own client code read it, move that read to the server. secure applies in every environment: browsers accept it from http://localhost, but a plain-HTTP development server on another host (a LAN address, for example) will no longer keep the session.
  • A session is only reused for the user it was minted for. The cookie now records the issuer and subject (iss and sub) of the OIDC token it was minted from, and protectMiddleware and protectClerkMiddleware replace a session left behind by another user (for example after a Clerk multi-session setActive switch, or a sign-out on a route the middleware does not match). If a replacement cannot be minted, the stale cookie is cleared. Cookies written by earlier versions carry no owner and are re-minted once. Ownership is read without verifying a signature, so the token you pass to protectMiddleware or getCtsToken must be one your application has already authenticated.
  • getCtsToken(oidcToken) only returns the cookie session when it belongs to the issuer and subject of oidcToken, and otherwise fetches a fresh session for that token.
  • getCtsToken() with no argument is deprecated. It returns the session cookie without checking who it belongs to, because it has no identity to compare it with, so pass the user's token whenever you have one. If a user signs out on a route the middleware does not match, call resetCtsToken there.
  • An expired or malformed session cookie no longer passes through or throws: it is treated as no session and re-minted or cleared by the middleware, instead of returning a 500 on every matched route.
  • A successful CTS response is now validated before it is used. A 2xx body that is not an unexpired token (non-JSON, the wrong shape, or an expiry already past) is reported as a failure instead of being written to the cookie or thrown out of the middleware.
  • GetCtsTokenResponse now uses success: true and success: false as literal discriminants, so checking result.success narrows result.ctsToken to a CtsToken.
  • When the middleware replaces or clears the session on a response that continues the request (NextResponse.next() or a rewrite), the request forwarded to route handlers and server components is updated in the same request, so getCtsToken() there never sees the previous user's session. Next.js does not do this for route handlers on any release tested, from 14.0.0 to 15.5.29. If the response you pass in already overrides the forwarded request headers, the forwarded session cookie is set to the one the middleware decided on, whether your override replaced the cookie or left it out. A session cookie your response already sets is likewise replaced with the kept session, or cleared when none was kept.
  • A CTS transport failure (DNS, TLS, connection reset) is reported as a failure instead of throwing, so the middleware still clears a stale session and does not return a 500. The error does not repeat the exception's message, which a fetch wrapper could fill with the request body.
  • A rejected Clerk getToken() no longer fails the request. It is treated as no token being available, so a stale session is still cleared.
  • On Next.js 14.2.8 to 14.2.23 and 15.0.x, Next.js itself sends the session token to the browser in an x-middleware-set-cookie response header whenever the session changes, so httpOnly does not protect it on those responses. Upgrade to at least 14.2.25 or 15.2.3, which also fix a critical Next.js middleware bypass (GHSA-f82v-jwr5-mffw).

• f7f1c76: The package's source links now point at its folder under
  languages/typescript/ in cipherstash/stack. The previous release linked to
  its old folder under packages/, which no longer exists, so the README's
  source links and the npm page's repository link returned 404.

@cipherstash/protect-ffi@0.33.1

Patch Changes

• 457b28e: The native binaries now use rustls 0.23.45, which fixes GHSA-2mjx-qc3c-rqvc
  (affecting rustls 0.23.13 up to 0.23.44). The package's API is unchanged.
• f7f1c76: The package's source links now point at its folder under
  languages/typescript/ in cipherstash/stack. The previous release linked to
  its old folder under packages/, which no longer exists, so the README's
  source links and the npm page's repository link returned 404.

@cipherstash/protect-ffi-darwin-arm64@0.33.1

Patch Changes

• 457b28e: The native binaries now use rustls 0.23.45, which fixes GHSA-2mjx-qc3c-rqvc
  (affecting rustls 0.23.13 up to 0.23.44). The package's API is unchanged.
• f7f1c76: The package's source links now point at its folder under
  languages/typescript/ in cipherstash/stack. The previous release linked to
  its old folder under packages/, which no longer exists, so the README's
  source links and the npm page's repository link returned 404.

@cipherstash/protect-ffi-darwin-x64@0.33.1

Patch Changes

• 457b28e: The native binaries now use rustls 0.23.45, which fixes GHSA-2mjx-qc3c-rqvc
  (affecting rustls 0.23.13 up to 0.23.44). The package's API is unchanged.
• f7f1c76: The package's source links now point at its folder under
  languages/typescript/ in cipherstash/stack. The previous release linked to
  its old folder under packages/, which no longer exists, so the README's
  source links and the npm page's repository link returned 404.

@cipherstash/protect-ffi-linux-arm64-gnu@0.33.1

Patch Changes

• 457b28e: The native binaries now use rustls 0.23.45, which fixes GHSA-2mjx-qc3c-rqvc
  (affecting rustls 0.23.13 up to 0.23.44). The package's API is unchanged.
• f7f1c76: The package's source links now point at its folder under
  languages/typescript/ in cipherstash/stack. The previous release linked to
  its old folder under packages/, which no longer exists, so the README's
  source links and the npm page's repository link returned 404.

@cipherstash/protect-ffi-linux-x64-gnu@0.33.1

Patch Changes

• 457b28e: The native binaries now use rustls 0.23.45, which fixes GHSA-2mjx-qc3c-rqvc
  (affecting rustls 0.23.13 up to 0.23.44). The package's API is unchanged.
• f7f1c76: The package's source links now point at its folder under
  languages/typescript/ in cipherstash/stack. The previous release linked to
  its old folder under packages/, which no longer exists, so the README's
  source links and the npm page's repository link returned 404.

@cipherstash/protect-ffi-linux-x64-musl@0.33.1

Patch Changes

• 457b28e: The native binaries now use rustls 0.23.45, which fixes GHSA-2mjx-qc3c-rqvc
  (affecting rustls 0.23.13 up to 0.23.44). The package's API is unchanged.
• f7f1c76: The package's source links now point at its folder under
  languages/typescript/ in cipherstash/stack. The previous release linked to
  its old folder under packages/, which no longer exists, so the README's
  source links and the npm page's repository link returned 404.

@cipherstash/protect-ffi-win32-x64-msvc@0.33.1

Patch Changes

• 457b28e: The native binaries now use rustls 0.23.45, which fixes GHSA-2mjx-qc3c-rqvc
  (affecting rustls 0.23.13 up to 0.23.44). The package's API is unchanged.
• f7f1c76: The package's source links now point at its folder under
  languages/typescript/ in cipherstash/stack. The previous release linked to
  its old folder under packages/, which no longer exists, so the README's
  source links and the npm page's repository link returned 404.

@cipherstash/stack-drizzle@1.3.0

Patch Changes

• Updated dependencies [ef3765b]
• Updated dependencies [43967d3]
• Updated dependencies [fd0ee91]
• Updated dependencies [4b14ee6]
• Updated dependencies [bbb1bfa]
• Updated dependencies [20ccab3]
  • @cipherstash/stack@1.3.0

@cipherstash/stack-supabase@1.3.0

Patch Changes

• 43967d3: Two validation errors no longer quote the caller's search term: a free-text search term that is too short, and a like/ilike pattern with wildcards the encrypted search cannot honour. Each now reports the term's length instead.

  The Supabase and DynamoDB adapters' error logs now record a failure's kind instead of its full message, which could quote a plaintext value. The kind is the error class plus any known encryption error type, FFI error code or auth code it carries; anything else is left out. The DynamoDB options.logger hook now receives a copy of the error whose message is that description, with the same code. The full message is still returned to the caller and passed to errorHandler. @cipherstash/stack/adapter-kit exports the shared describeFailure and likeWildcardError helpers that produce these messages.

• bbb1bfa: encrypt on a types.Json column now types its result as Encrypted | null when the plaintext may be null. A null document has always resolved to { data: null } at runtime (it is stored as SQL NULL, not encrypted), but the result was typed Encrypted, so result.data.c compiled and then threw. encrypt is now generic over the plaintext's type: a literal null, or a value typed JsonDocument, gives data: Encrypted | null, through .withLockContext() and .audit() too. A value TypeScript knows is non-null, and every scalar column, still gives data: Encrypted. No runtime change. If your code reads result.data after encrypting a Json value that may be null, it must now check for null first. EncryptOperation and EncryptOperationWithLockContext take an optional result type parameter (default Encrypted), and the EncryptResult<P> helper type is exported from @cipherstash/stack/encryption.

  @cipherstash/stack-supabase: the per-term query-encryption fallback (used when the client has no bulkEncrypt) now rejects a null envelope instead of sending the string "null" as a filter operand, matching the bulk path.

  stash: the stash-encryption skill now says what encrypt returns for a null Json document.

• ab24980: Reading a table that has no encrypted columns no longer fails. Such a table is
  not registered with the encryption client, and the adapter still asked the
  client to decrypt its rows, so every select() on it, and every insert, update
  or upsert followed by .select(), returned "decryptModel received a table this
  client was not initialized with" (or threw, with throwOnError). On the
  wasm-inline entry, writes to such a table failed the same way. The adapter now
  passes the rows of a table with no encrypted columns straight through, in both
  introspection and declared mode, without calling the encryption client.

  A row from such a table that still holds an encrypted value, from an embedded
  relation to an encrypted table or from an undeclared eql_v3_* column in
  declared mode, is refused with an error naming the column, rather than returned
  as ciphertext.

• Updated dependencies [ef3765b]

• Updated dependencies [43967d3]

• Updated dependencies [fd0ee91]

• Updated dependencies [4b14ee6]

• Updated dependencies [bbb1bfa]

• Updated dependencies [20ccab3]

  • @cipherstash/stack@1.3.0

@cipherstash/wizard@1.3.0

Patch Changes

• 472cbf0: Add @modelcontextprotocol/sdk ^1.32.0 as a direct dependency so the
  auto-installed peer of @anthropic-ai/claude-agent-sdk resolves to a release
  patched against GHSA-6qxp-vccf-f47h / CVE-2026-104850, instead of the vulnerable
  1.29.0 previously locked. The wizard never imports the MCP SDK directly — this
  is a peer-resolution pin only; no API change.
• Updated dependencies [db76cf4]
• Updated dependencies [a1765f1]
• Updated dependencies [fb75f40]
• Updated dependencies [a076649]
• Updated dependencies [457b28e]
• Updated dependencies [b480735]
• Updated dependencies [fd0ee91]
  • @cipherstash/auth@1.0.0

@cipherstash/auth-darwin-arm64@1.0.0

@cipherstash/auth-darwin-x64@1.0.0

@cipherstash/auth-linux-arm64-gnu@1.0.0

@cipherstash/auth-linux-x64-gnu@1.0.0

@cipherstash/auth-linux-x64-musl@1.0.0

@cipherstash/auth-win32-x64-msvc@1.0.0

@cipherstash/e2e@0.0.8

Patch Changes

• Updated dependencies [ef3765b]
• Updated dependencies [db76cf4]
• Updated dependencies [c2798f0]
• Updated dependencies [bde5b73]
• Updated dependencies [b15c1ad]
• Updated dependencies [169f985]
• Updated dependencies [587d06e]
• Updated dependencies [bd7e55f]
• Updated dependencies [e5a3575]
• Updated dependencies [43967d3]
• Updated dependencies [fd0ee91]
• Updated dependencies [ae63dcd]
• Updated dependencies [4b14ee6]
• Updated dependencies [bbb1bfa]
• Updated dependencies [20ccab3]
• Updated dependencies [bca02b6]
• Updated dependencies [8280acf]
• Updated dependencies [9a7815d]
• Updated dependencies [ef3765b]
• Updated dependencies [472cbf0]
  • @cipherstash/stack@1.3.0
  • stash@1.3.0
  • @cipherstash/wizard@1.3.0

@cipherstash/basic-example@1.2.19

Patch Changes

• Updated dependencies [ef3765b]
• Updated dependencies [43967d3]
• Updated dependencies [fd0ee91]
• Updated dependencies [4b14ee6]
• Updated dependencies [bbb1bfa]
• Updated dependencies [20ccab3]
  • @cipherstash/stack@1.3.0
  • @cipherstash/stack-drizzle@1.3.0

@cipherstash/prisma-example@0.1.5

Patch Changes

• Updated dependencies [ef3765b]
• Updated dependencies [ae63dcd]
• Updated dependencies [43967d3]
• Updated dependencies [fd0ee91]
• Updated dependencies [4b14ee6]
• Updated dependencies [bbb1bfa]
• Updated dependencies [20ccab3]
  • @cipherstash/stack@1.3.0
  • @cipherstash/stack-prisma@1.3.0

@cipherstash/bench@0.0.10

Patch Changes

• Updated dependencies [ef3765b]
• Updated dependencies [43967d3]
• Updated dependencies [fd0ee91]
• Updated dependencies [4b14ee6]
• Updated dependencies [bbb1bfa]
• Updated dependencies [20ccab3]
  • @cipherstash/stack@1.3.0
  • @cipherstash/stack-drizzle@1.3.0

@cipherstash/ffi-integration-tests@1.0.3

Patch Changes

• Updated dependencies [db76cf4]
• Updated dependencies [a1765f1]
• Updated dependencies [fb75f40]
• Updated dependencies [a076649]
• Updated dependencies [457b28e]
• Updated dependencies [b480735]
• Updated dependencies [fd0ee91]
• Updated dependencies [457b28e]
• Updated dependencies [f7f1c76]
  • @cipherstash/auth@1.0.0
  • @cipherstash/protect-ffi@0.33.1

@cipherstash/test-kit@0.0.6

Patch Changes

• Updated dependencies [ef3765b]
• Updated dependencies [43967d3]
• Updated dependencies [fd0ee91]
• Updated dependencies [4b14ee6]
• Updated dependencies [bbb1bfa]
• Updated dependencies [20ccab3]
  • @cipherstash/stack@1.3.0

[auxesis]

Fix for failing build is #1081.

[coderabbitai]

Important

Review skipped

Bot user detected.

To trigger a single review, invoke the @coderabbitai review command.

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI; Enforced overrides: Organization
• Review profile: CHILL
• Plan: Advanced
• Run ID: dcd33fc7-ca23-4ec2-88e9-87d5fc268039

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• Trigger review