deepagents==0.7.25

langchain-ai/deepagentsdeepagents==0.7.25Oct 14, 2026by github-actions[bot]

Upgrading?

AI Summary

This release introduces support for reading inline skill tools from model profiles and includes critical security and performance fixes for the sandbox environment.

Key Highlights

  • Support for reading inline skill tool support from model profiles.
  • Security fix: Filesystem permission checks now verify symlink-resolved paths to prevent access bypassing.
  • Performance improvement: Sandbox searches now prefer ripgrep.
  • Performance improvement: Sandbox grep skips .git directories.
  • Bug fix: Fixed path-glob grep in sandboxes.

New Features

  • Read inline skill tool support from model profiles.

Full Release Notes

### Features

- Read inline skill tool support from model profiles. ([#6879](https://github.com/langchain-ai/deepagents/pull/6879))

### Bug Fixes

- Filesystem permission rules and `interrupt_on` approvals now check both the requested and symlink-resolved paths; previously, `FilesystemBackend` followed symlinks after the check, potentially allowing access to denied paths. Review your rules and approvals for symlink targets. ([#6818](https://github.com/langchain-ai/deepagents/pull/6818))
- Fixed path-glob `grep` in sandboxes. ([#6910](https://github.com/langchain-ai/deepagents/pull/6910))

### Performance Improvements

- Sandbox searches now prefer ripgrep, and sandbox `grep` skips `.git`. ([#6919](https://github.com/langchain-ai/deepagents/pull/6919), [#6910](https://github.com/langchain-ai/deepagents/pull/6910))

---

**Special thanks** to everyone who reported the issues addressed in this release: @AlcebiadesFilho ([#6860](https://github.com/langchain-ai/deepagents/issues/6860))

Internal maintainers: @langchain-infra, @mdrxy, @richardscarrott, @thushanth-bengre-langchain

Released by: @hntrl

<details>
<summary>Git log since deepagents==0.7.24</summary>

This commit history includes changes to this package. Commits are listed newest first.

- [`c6c340d`](https://github.com/langchain-ai/deepagents/commit/c6c340d47d2013a2e7b7230698f73977b7de6756) release(deepagents): 0.7.25 (#6901)
- [`f5d9c31`](https://github.com/langchain-ai/deepagents/commit/f5d9c31163dda54898bdb4f73a2580c9430a16d5) feat(sdk): read inline skill tool support from model profiles (#6879)
- [`7a655f7`](https://github.com/langchain-ai/deepagents/commit/7a655f77382484a6e6ca27e8724b08064af670c8) perf(sdk): prefer ripgrep for sandbox searches (#6919)
- [`f8ecffa`](https://github.com/langchain-ai/deepagents/commit/f8ecffad6523775a323a807bc57097a98aa27e01) perf(sdk): skip `.git` in sandbox `grep` and fix path-glob `grep` (#6910)
- [`c96d912`](https://github.com/langchain-ai/deepagents/commit/c96d912ba70a601f0e259cded6b428f957b88cbf) chore(deps): raise dependency minimums for `all` (#6915)
- [`7a69743`](https://github.com/langchain-ai/deepagents/commit/7a6974309ac78f0b2ea636afc70bd8a0268bed7f) fix(sdk): check filesystem permissions against the symlink-resolved path (#6818)
- [`53238e9`](https://github.com/langchain-ai/deepagents/commit/53238e943fb2edd25c157178c4d6e73e2ba75f2c) chore(deps): raise dependency minimums for `all` (#6899)
- [`d312bcd`](https://github.com/langchain-ai/deepagents/commit/d312bcda3c71dfc4201b993776a413c8b457b483) chore(deps): raise dependency minimums for `all` (#6898)

</details>

Related Repositories