Skip to content

fix(agent): restrict background review fork to memory + skills toolsets#16569

Merged
teknium1 merged 1 commit into
mainfrom
hermes/hermes-6e833b3d
Apr 27, 2026
Merged

fix(agent): restrict background review fork to memory + skills toolsets#16569
teknium1 merged 1 commit into
mainfrom
hermes/hermes-6e833b3d

Conversation

@teknium1

Copy link
Copy Markdown
Contributor

Summary

Background memory/skill review fork can no longer run terminal, send_message, delegate_task, browser, web, or file tools. Restricted to memory + skills only — which is everything the review prompts actually need.

Fixes #15204.

Changes

  • run_agent.py: one-line enabled_toolsets=["memory", "skills"] added to the AIAgent(...) construction in _spawn_background_review().
  • tests/run_agent/test_background_review_toolset_restriction.py: regression coverage.

Validation

Targeted test suite: 2/2 passed.

Credit

Salvage of #16001 by @luyao618 onto current main (251 commits ahead of the original branch). Cherry-picked with original authorship preserved.

…lsets

The background skill/memory review agent was created without toolset
restrictions, inheriting the full default tool set. This allowed it to
use terminal, send_message, delegate_task, and other tools outside its
intended scope, potentially performing unrelated side effects after
skill creation.

Restrict the review agent to only memory and skills toolsets by passing
enabled_toolsets=['memory', 'skills'] during AIAgent construction.

Fixes #15204
@teknium1 teknium1 merged commit 8ad29a9 into main Apr 27, 2026
11 of 12 checks passed
@teknium1 teknium1 deleted the hermes/hermes-6e833b3d branch April 27, 2026 13:41
@alt-glitch alt-glitch added type/security Security vulnerability or hardening P2 Medium — degraded but workaround exists comp/agent Core agent loop, run_agent.py, prompt builder labels Apr 27, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

comp/agent Core agent loop, run_agent.py, prompt builder P2 Medium — degraded but workaround exists type/security Security vulnerability or hardening

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Background skill-review agent can perform non-skill side effects after creating a skill

3 participants