Skip to content

chore(deps-dev): Bump @ai-hero/sandcastle from 0.12.0 to 0.14.0 in the all-npm group - #344

Open
dependabot[bot] wants to merge 1 commit into
mainfrom
dependabot/npm_and_yarn/main/all-npm-2baaaa7ded
Open

dependabot[bot] wants to merge 1 commit into
mainfrom
dependabot/npm_and_yarn/main/all-npm-2baaaa7ded

Conversation

@dependabot

@dependabot dependabot Bot commented on behalf of github Nov 15, 2026

Copy link
Copy Markdown
Contributor

Bumps the all-npm group with 1 update: @ai-hero/sandcastle.

Updates @ai-hero/sandcastle from 0.12.0 to 0.14.0

Release notes

Sourced from @​ai-hero/sandcastle's releases.

v0.14.0

Minor Changes

  • d99fff8: Add an E2B isolated sandbox provider, e2b() from @ai-hero/sandcastle/sandboxes/e2b. It needs the optional e2b peer dependency and E2B_API_KEY (or the apiKey option). Use a template that ships your agent's CLI, such as e2b({ template: "claude" }).
  • 8783573: Add modal(), an isolated sandbox provider for Modal sandboxes, at @ai-hero/sandcastle/sandboxes/modal. Install modal (an optional peer dependency) and set MODAL_TOKEN_ID and MODAL_TOKEN_SECRET, or pass tokenId and tokenSecret. By default it builds a node:24-bookworm image with Claude Code installed; pass image for your own. Sandboxes live for an hour unless timeoutMs says otherwise, and sandboxId is the Modal sandbox id.
  • 775a8df: Add sprites(), a Fly.io Sprites isolated sandbox provider, at @ai-hero/sandcastle/sandboxes/sprites. It needs @fly/sprites (an optional peer dependency, which asks for Node 24 or later) and a token in SPRITES_TOKEN (or SPRITE_TOKEN, or the token option). Each sandbox is a new Sprite with the repo at /home/sprite/workspace; sandboxId is the Sprite's name. Sprites persist until deleted, so Sandcastle deletes the Sprite on close(), when setup fails, and from a background process when the run is interrupted or exits without closing it.

Patch Changes

  • 73ce215: docker() and podman(): the worktree admin directories can no longer be renamed from inside the sandbox.
  • 14507d6: Add a Glossary page to the user docs (docs/guide/glossary.md, in the Reference group) that defines every term in one place. Guide pages no longer open with definition blocks; instead, each page links the first mention of a term to its glossary entry.

v0.13.0

Highlights

  • Native sandboxes. bwrap() (Linux and WSL2) and seatbelt() (macOS) run the agent in an OS sandbox with no container and no image. They are now the documented default. sandcastle init --sandbox bwrap scaffolds a project that needs no Dockerfile.
  • Resume for more agents. Copilot, Cursor and OpenCode sessions can now be resumed; OpenCode can also fork.
  • sandboxId on the sandbox handle (the container name on Docker and Podman, the sandbox's name or id on Vercel and Daytona), and capDrop for Docker and Podman.
  • Security fixes. The host's .git is mounted read-only in bind-mount sandboxes, and host-side git ignores hooks and config that run programs. Sync-out refuses paths that leave the worktree, and fetch only uses allowed transports. See the entries below.

Behaviour changes you may notice

  • git config without --global fails inside a bind-mount sandbox (the repo's git config is read-only there).
  • Host-side merges no longer run the repo's git hooks.
  • Sync-out now fails with SyncOutPathError where it used to skip a file.

Minor Changes

  • 0210ba0: Add session resume to the copilot() agent. Sandcastle now captures each Copilot session's events.jsonl to ~/.copilot/session-state/<id>/ on the host, and r.resume() (or resumeSession) copies it back into the sandbox and runs copilot --resume=<id>. Opt out with copilot(model, { captureSessions: false }). The Copilot CLI has no fork, so r.fork() throws a ForkNotSupportedError naming copilot before any sandbox starts.

  • cd39292: Add session resume to the cursor() agent. Sandcastle now captures each Cursor chat's store.db to ~/.cursor/chats/<md5-of-cwd>/<chat-id>/ on the host, byte for byte, and r.resume() (or resumeSession) copies it back into the sandbox and runs agent --resume=<id>. Opt out with cursor(model, { captureSessions: false }). The Cursor CLI has no fork, so r.fork() throws a ForkNotSupportedError naming cursor before any sandbox starts.

  • 734adb6: Fix the Context window: NNNk line so it shows the last model call's input tokens (cache included) for every agent, never a running total. claudeCode now reads it from the stream, so it also shows on noSandbox, vercel and with captureSessions: false. codex reads it from its captured session instead of the cumulative turn.completed total, and shows the model's limit (34k / 258k). pi, opencode and copilot now show it. cursor shows none, because it reports only a whole-turn total. When an iteration fails, the line still shows for the iterations that finished. New IterationResult.contextWindow ({ tokens, limit? }).

  • 2472e66: sandcastle init with --sandbox bwrap or --sandbox seatbelt now checks the host: it warns, with the exact install command, when the sandbox tool or the chosen agent's CLI is missing, and still writes the files. With the beads issue tracker it scaffolds an identity mount of .beads/ (at its host path, with BEADS_DIR set to match) and skips the bd version warning, since the host's own bd is used. With GitHub Issues, the next steps say to put GH_TOKEN in .sandcastle/.env.

  • 5d4d06b: sandcastle init offers the native sandbox providers. The interactive prompt lists the host OS's native provider first and preselects it, marked "(recommended)": bwrap on Linux (and WSL2), seatbelt on macOS, then Docker and Podman. --sandbox bwrap|seatbelt|docker|podman is accepted on any OS. A native provider scaffolds no Dockerfile or Containerfile and has no build-image prompt, --build-image flag or build-image next step; the main file imports the chosen provider. Non-interactive init still requires --sandbox, and init --sandbox docker is unchanged.

  • de65fb3: Add bwrap(), a native sandbox provider for Linux and WSL2, at @ai-hero/sandcastle/sandboxes/bwrap. It runs each command in Bubblewrap, with no image and no daemon: the host's system dirs read-only, the worktree at its host path, a private home at /home/agent and a private /tmp (both removed on close), its own PID namespace (every process the command started dies with it, even when your Node process is hard-killed: SIGKILL, the OOM killer), the host network, and only the env vars Sandcastle passes in. exec with sudo: true rejects, and interactive() with a native provider throws NativeInteractiveNotSupportedError.

    Constructing bwrap() checks once per process that Bubblewrap is usable, and throws a new NativeSandboxUnavailableError (exported from @ai-hero/sandcastle, with message and an optional fixCommand) if it isn't: on a non-Linux OS (use seatbelt() on macOS), when bwrap isn't on PATH (the fix is your distro's install command), or when the kernel blocks unprivileged user namespaces (Ubuntu 24.04+ AppArmor; the fix is an AppArmor profile for bwrap). It throws before any worktree exists and never falls back to running unsandboxed.

    bwrap() runs your host's agent CLI with no image: every directory on the host PATH is readable inside the sandbox, plus the install root of each executable symlinked from one (an nvm-installed codex, a claude in ~/.local/bin), and the child's PATH is the host PATH filtered to what it can see. The install root is the executable's package when it sits under a node_modules dir, and otherwise only the executable's own directory, so an npm linked checkout or a stray package.json above your projects doesn't expose them. A PATH entry equal to your home directory or an ancestor of it, and an install root holding a .git, are never mounted. seatbelt() resolves its toolchain by the same rules. Add other paths with the mounts option: each appears at its own host path, so sandboxPath must equal hostPath (bwrap() throws otherwise), and readonly defaults to false.

  • 6a12a94: Add seatbelt(), a native sandbox provider for macOS 14 or later, at @ai-hero/sandcastle/sandboxes/seatbelt. It runs each command under the system /usr/bin/sandbox-exec with a deny-by-default Seatbelt profile, with no image and no daemon: the host's system toolchain and PATH dirs read-only, the worktree and git dirs read/write at their host paths, a private home and temp dir (both removed on close), the host network, and only the env vars Sandcastle passes in plus PATH, HOME and TMPDIR. The git control files stay read-only and the dirs holding them can't be renamed aside. The agent can't signal your processes, though Seatbelt can't fully hide them. Each exec leads its own process group, which is killed when it exits, on close, on shutdown and, through a small host watchdog, when your Node process is hard-killed (SIGKILL, the OOM killer). A known limit: a process that leaves its group with setsid or setpgid survives all of these, still under the Seatbelt profile. Copies to and from /home/agent land in the private home. Session capture and resume work for every agent, including opencode session export and Claude Code subagent transcripts. Add other paths with mounts (each at its own host path, as for bwrap()). exec with sudo: true rejects, and interactive() throws NativeInteractiveNotSupportedError.

    Constructing seatbelt() checks once per process that Seatbelt works, and throws NativeSandboxUnavailableError if it doesn't: on a non-macOS OS (use bwrap() on Linux or WSL2), on macOS older than 14, or when sandbox-exec is missing or fails (for instance inside another Seatbelt sandbox). There is no install command on macOS, so the error has no fixCommand and points to docker(). It never falls back to running unsandboxed.

  • 57faed6: Add session resume and fork to the opencode() agent. OpenCode keeps sessions in its SQLite database, so after each run Sandcastle runs opencode export <id> in the sandbox and stores the JSON under ~/.local/share/opencode/sandcastle-sessions/<id>.json on the host. r.resume() (or resumeSession) copies it back, runs opencode import, and continues with opencode run --session <id>; r.fork() adds --fork, so the fork gets a new session id and the parent is left alone. Opt out with opencode(model, { captureSessions: false }).

  • e9b1ce1: A BindMountSandboxHandle can now report sandboxHomedir and sandboxTmpdir: the agent's home and a writable temp dir, as commands inside the sandbox see them. Session capture and resume build every path they use inside the sandbox from these (the Claude Code projects dir and its subagent transcripts, the Codex, pi, Cursor and Copilot session dirs, and the opencode export file) instead of a fixed /home/agent and /tmp. Both are optional and default to /home/agent and /tmp, so docker(), podman() and custom providers behave as before, and an explicit sessionStorage sandbox dir still wins.

Patch Changes

... (truncated)

Changelog

Sourced from @​ai-hero/sandcastle's changelog.

0.14.0

Minor Changes

  • d99fff8: Add an E2B isolated sandbox provider, e2b() from @ai-hero/sandcastle/sandboxes/e2b. It needs the optional e2b peer dependency and E2B_API_KEY (or the apiKey option). Use a template that ships your agent's CLI, such as e2b({ template: "claude" }).
  • 8783573: Add modal(), an isolated sandbox provider for Modal sandboxes, at @ai-hero/sandcastle/sandboxes/modal. Install modal (an optional peer dependency) and set MODAL_TOKEN_ID and MODAL_TOKEN_SECRET, or pass tokenId and tokenSecret. By default it builds a node:24-bookworm image with Claude Code installed; pass image for your own. Sandboxes live for an hour unless timeoutMs says otherwise, and sandboxId is the Modal sandbox id.
  • 775a8df: Add sprites(), a Fly.io Sprites isolated sandbox provider, at @ai-hero/sandcastle/sandboxes/sprites. It needs @fly/sprites (an optional peer dependency, which asks for Node 24 or later) and a token in SPRITES_TOKEN (or SPRITE_TOKEN, or the token option). Each sandbox is a new Sprite with the repo at /home/sprite/workspace; sandboxId is the Sprite's name. Sprites persist until deleted, so Sandcastle deletes the Sprite on close(), when setup fails, and from a background process when the run is interrupted or exits without closing it.

Patch Changes

  • 73ce215: docker() and podman(): the worktree admin directories can no longer be renamed from inside the sandbox.
  • 14507d6: Add a Glossary page to the user docs (docs/guide/glossary.md, in the Reference group) that defines every term in one place. Guide pages no longer open with definition blocks; instead, each page links the first mention of a term to its glossary entry.

0.13.0

Minor Changes

  • 0210ba0: Add session resume to the copilot() agent. Sandcastle now captures each Copilot session's events.jsonl to ~/.copilot/session-state/<id>/ on the host, and r.resume() (or resumeSession) copies it back into the sandbox and runs copilot --resume=<id>. Opt out with copilot(model, { captureSessions: false }). The Copilot CLI has no fork, so r.fork() throws a ForkNotSupportedError naming copilot before any sandbox starts.

  • cd39292: Add session resume to the cursor() agent. Sandcastle now captures each Cursor chat's store.db to ~/.cursor/chats/<md5-of-cwd>/<chat-id>/ on the host, byte for byte, and r.resume() (or resumeSession) copies it back into the sandbox and runs agent --resume=<id>. Opt out with cursor(model, { captureSessions: false }). The Cursor CLI has no fork, so r.fork() throws a ForkNotSupportedError naming cursor before any sandbox starts.

  • 734adb6: Fix the Context window: NNNk line so it shows the last model call's input tokens (cache included) for every agent, never a running total. claudeCode now reads it from the stream, so it also shows on noSandbox, vercel and with captureSessions: false. codex reads it from its captured session instead of the cumulative turn.completed total, and shows the model's limit (34k / 258k). pi, opencode and copilot now show it. cursor shows none, because it reports only a whole-turn total. When an iteration fails, the line still shows for the iterations that finished. New IterationResult.contextWindow ({ tokens, limit? }).

  • 2472e66: sandcastle init with --sandbox bwrap or --sandbox seatbelt now checks the host: it warns, with the exact install command, when the sandbox tool or the chosen agent's CLI is missing, and still writes the files. With the beads issue tracker it scaffolds an identity mount of .beads/ (at its host path, with BEADS_DIR set to match) and skips the bd version warning, since the host's own bd is used. With GitHub Issues, the next steps say to put GH_TOKEN in .sandcastle/.env.

  • 5d4d06b: sandcastle init offers the native sandbox providers. The interactive prompt lists the host OS's native provider first and preselects it, marked "(recommended)": bwrap on Linux (and WSL2), seatbelt on macOS, then Docker and Podman. --sandbox bwrap|seatbelt|docker|podman is accepted on any OS. A native provider scaffolds no Dockerfile or Containerfile and has no build-image prompt, --build-image flag or build-image next step; the main file imports the chosen provider. Non-interactive init still requires --sandbox, and init --sandbox docker is unchanged.

  • de65fb3: Add bwrap(), a native sandbox provider for Linux and WSL2, at @ai-hero/sandcastle/sandboxes/bwrap. It runs each command in Bubblewrap, with no image and no daemon: the host's system dirs read-only, the worktree at its host path, a private home at /home/agent and a private /tmp (both removed on close), its own PID namespace (every process the command started dies with it, even when your Node process is hard-killed: SIGKILL, the OOM killer), the host network, and only the env vars Sandcastle passes in. exec with sudo: true rejects, and interactive() with a native provider throws NativeInteractiveNotSupportedError.

    Constructing bwrap() checks once per process that Bubblewrap is usable, and throws a new NativeSandboxUnavailableError (exported from @ai-hero/sandcastle, with message and an optional fixCommand) if it isn't: on a non-Linux OS (use seatbelt() on macOS), when bwrap isn't on PATH (the fix is your distro's install command), or when the kernel blocks unprivileged user namespaces (Ubuntu 24.04+ AppArmor; the fix is an AppArmor profile for bwrap). It throws before any worktree exists and never falls back to running unsandboxed.

    bwrap() runs your host's agent CLI with no image: every directory on the host PATH is readable inside the sandbox, plus the install root of each executable symlinked from one (an nvm-installed codex, a claude in ~/.local/bin), and the child's PATH is the host PATH filtered to what it can see. The install root is the executable's package when it sits under a node_modules dir, and otherwise only the executable's own directory, so an npm linked checkout or a stray package.json above your projects doesn't expose them. A PATH entry equal to your home directory or an ancestor of it, and an install root holding a .git, are never mounted. seatbelt() resolves its toolchain by the same rules. Add other paths with the mounts option: each appears at its own host path, so sandboxPath must equal hostPath (bwrap() throws otherwise), and readonly defaults to false.

  • 6a12a94: Add seatbelt(), a native sandbox provider for macOS 14 or later, at @ai-hero/sandcastle/sandboxes/seatbelt. It runs each command under the system /usr/bin/sandbox-exec with a deny-by-default Seatbelt profile, with no image and no daemon: the host's system toolchain and PATH dirs read-only, the worktree and git dirs read/write at their host paths, a private home and temp dir (both removed on close), the host network, and only the env vars Sandcastle passes in plus PATH, HOME and TMPDIR. The git control files stay read-only and the dirs holding them can't be renamed aside. The agent can't signal your processes, though Seatbelt can't fully hide them. Each exec leads its own process group, which is killed when it exits, on close, on shutdown and, through a small host watchdog, when your Node process is hard-killed (SIGKILL, the OOM killer). A known limit: a process that leaves its group with setsid or setpgid survives all of these, still under the Seatbelt profile. Copies to and from /home/agent land in the private home. Session capture and resume work for every agent, including opencode session export and Claude Code subagent transcripts. Add other paths with mounts (each at its own host path, as for bwrap()). exec with sudo: true rejects, and interactive() throws NativeInteractiveNotSupportedError.

    Constructing seatbelt() checks once per process that Seatbelt works, and throws NativeSandboxUnavailableError if it doesn't: on a non-macOS OS (use bwrap() on Linux or WSL2), on macOS older than 14, or when sandbox-exec is missing or fails (for instance inside another Seatbelt sandbox). There is no install command on macOS, so the error has no fixCommand and points to docker(). It never falls back to running unsandboxed.

  • 57faed6: Add session resume and fork to the opencode() agent. OpenCode keeps sessions in its SQLite database, so after each run Sandcastle runs opencode export <id> in the sandbox and stores the JSON under ~/.local/share/opencode/sandcastle-sessions/<id>.json on the host. r.resume() (or resumeSession) copies it back, runs opencode import, and continues with opencode run --session <id>; r.fork() adds --fork, so the fork gets a new session id and the parent is left alone. Opt out with opencode(model, { captureSessions: false }).

  • e9b1ce1: A BindMountSandboxHandle can now report sandboxHomedir and sandboxTmpdir: the agent's home and a writable temp dir, as commands inside the sandbox see them. Session capture and resume build every path they use inside the sandbox from these (the Claude Code projects dir and its subagent transcripts, the Codex, pi, Cursor and Copilot session dirs, and the opencode export file) instead of a fixed /home/agent and /tmp. Both are optional and default to /home/agent and /tmp, so docker(), podman() and custom providers behave as before, and an explicit sessionStorage sandbox dir still wins.

Patch Changes

  • 94c8953: sandcastle init --issue-tracker beads now checks the host's bd version against the image's pinned bd 1.3.2 and warns on a mismatch. Every beads sandbox opens the host's .beads/ with the image's bd, which migrates the database; an older host bd then fails to create issues (fk_events_issue). The warning names both versions and says which one to change.
  • b6cb2bb: Fix the beads issue tracker in worktree sandboxes. sandcastle init --issue-tracker beads now scaffolds a sandboxOptions that every sandbox call receives: a mount of the host's .beads/ and BEADS_DIR pointing at it, so merge-to-head and branch sandboxes share the host's beads database instead of failing with "no beads database found". The beads image now installs a pinned, checksum-verified bd release (ARG BEADS_VERSION, default 1.3.2) instead of the unpinned main-branch install script.
  • 912b7ef: bwrap() now removes each sandbox's private home and /tmp when a run is interrupted. Before, SIGINT or SIGTERM (a Ctrl-C) killed the sandbox's processes but left a sandcastle-bwrap-* dir in the host's temp dir, holding the agent's home (.claude/, .claude.json, .gitconfig), one more after every interrupt. The same removal runs when the process exits without closing the sandbox.
  • 00cfde0: Add a capDrop option to docker() and podman(). Each listed Linux capability (named without the CAP_ prefix, or "ALL") is passed to docker run/podman run as its own --cap-drop flag, so a sandbox can run without capabilities such as NET_RAW. The names are typed as LinuxCapability, so a misspelt one fails at compile time.
  • a9a04cc: Fix Claude Code session capture for repo paths containing . or other non-alphanumeric characters. encodeProjectPath now names the ~/.claude/projects/ directory exactly as Claude Code does: every character that isn't an ASCII letter or digit becomes - (including ., _, spaces and a Windows drive colon), and names over 200 characters are truncated with Claude Code's hash suffix. Previously only path separators were replaced, so a captured session for a repo like ~/repos/my.app landed in a directory claude --resume never looks in, and native sandboxes failed session capture for every worktree under .sandcastle/worktrees.
  • 516a8d0: Fix copyToWorktree so each path lands at the same relative path in the worktree: a directory that already exists there (e.g. one holding tracked files) is merged into instead of nested (cfg/cfg), and missing parent directories of a nested path (e.g. deep/a/b/c.txt under an ignored deep/) are created instead of failing the run.
  • 26ec236: createSandbox() and worktree.createSandbox() now pass the agent provider's env (e.g. claudeCode(model, { env })) to the agent on each sandbox.run(). It was silently dropped, because the container starts before an agent is chosen. The env is exported for that run's agent process only, and keys that overlap the sandbox provider's env are rejected, as in run().
  • 21e4c55: Fix {{TARGET_BRANCH}} under createWorktree(): wt.run() and wt.interactive() now set it to the host's current branch, as run() and interactive() do, instead of the worktree's own branch.
  • 43a38c8: Show cursor's tool calls in the run log again. Current Cursor CLIs report shell commands as shellToolCall and file writes as editToolCall, which the parser didn't know, so a cursor run showed prose only. They now display as Bash(<command>) and Edit(<path>).
  • bf39c6f: Restructure the user docs in docs/guide/ around the glossary: one page per major term (core concepts, sandbox providers, branch strategies, agent providers, agent sessions, iterations, structured output, prompts, hooks, init, run logs), each opening with the term's definition, plus a Reference group for run() and the CLI. The glossary is now GLOSSARY.md (was CONTEXT.md).
  • 5ec951f: Move the user docs out of README.md into one page per topic under docs/guide/, ordered by docs/guide/toc.json. The README keeps the overview and links to every page; the same pages are published at aihero.dev/sandcastle.
  • 5999abe: Fix the Codex .env.example template written by sandcastle init: it now uses CODEX_API_KEY (the only API-key variable codex exec reads from the environment) instead of OPENAI_KEY, which nothing read.
  • 8adb685: Make daytona() work on Daytona's default snapshot. The repo now syncs into <workdir>/workspace, which the daytona user can create, instead of replacing the home directory. The resolved env (.sandcastle/.env keys, daytona({ env }), the agent's env) now reaches the sandbox as envVars, with explicit create.envVars taking precedence. The @daytona/sdk peer range is widened to >=0.164.0 <1.0.0.

... (truncated)

Commits
  • 7fe4a91 Version Packages (#1472)
  • f50781b chore(lint): split getNextStepsLines under max-lines-per-function (#1489)
  • 3c9d7be chore(lint): split run() under the function-size rules (#1491)
  • 4be344a chore(lint): share createSandbox's start, hook and handle helpers (#1490)
  • 98c7b61 chore(lint): void returns and function size in daytona (#1482)
  • f183746 chore(lint): no-conditional-in-test in createSandbox.test.ts (#1484)
  • 70725ed chore(lint): prefer-destructuring in InitService (#1480)
  • 58c4178 chore(lint): no-unused-vars and prefer-strict-equal in InitService.test.ts (#...
  • f834a89 chore(lint): split fetchPullRequestContext under max-lines-per-function (#1485)
  • 6231c1d chore(lint): strict-void-return in .sandcastle/run.ts (#1483)
  • Additional commits viewable in compare view

Dependabot compatibility score

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.


Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

  • @dependabot rebase will rebase this PR
  • @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
  • @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
  • @dependabot ignore <dependency name> major version will close this group update PR and stop Dependabot creating any more for the specific dependency's major version (unless you unignore this specific dependency's major version or upgrade to it yourself)
  • @dependabot ignore <dependency name> minor version will close this group update PR and stop Dependabot creating any more for the specific dependency's minor version (unless you unignore this specific dependency's minor version or upgrade to it yourself)
  • @dependabot ignore <dependency name> will close this group update PR and stop Dependabot creating any more for the specific dependency (unless you unignore this specific dependency or upgrade to it yourself)
  • @dependabot unignore <dependency name> will remove all of the ignore conditions of the specified dependency
  • @dependabot unignore <dependency name> <ignore condition> will remove the ignore condition of the specified dependency and ignore conditions

Bumps the all-npm group with 1 update: [@ai-hero/sandcastle](https://github.com/mattpocock/sandcastle).


Updates `@ai-hero/sandcastle` from 0.12.0 to 0.14.0
- [Release notes](https://github.com/mattpocock/sandcastle/releases)
- [Changelog](https://github.com/mattpocock/sandcastle/blob/main/CHANGELOG.md)
- [Commits](mattpocock/sandcastle@v0.12.0...v0.14.0)

---
updated-dependencies:
- dependency-name: "@ai-hero/sandcastle"
  dependency-version: 0.14.0
  dependency-type: direct:development
  update-type: version-update:semver-minor
  dependency-group: all-npm
...

Signed-off-by: dependabot[bot] <support@github.com>
@dependabot dependabot Bot added dependencies Pull requests that update a dependency npm Pull requests that update npm packages labels Nov 15, 2026
@github-actions
github-actions Bot enabled auto-merge (squash) November 15, 2026 16:12

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

dependencies Pull requests that update a dependency npm Pull requests that update npm packages

Projects

None yet

Development

Successfully merging this pull request may close these issues.

0 participants