Skip to content

test(team): M15 real-model Team mode security suite - #4304

Merged
DoodleBears merged 10 commits into
mainfrom
claude/4197-team-security-suite
Nov 17, 2026
Merged

DoodleBears merged 10 commits into
mainfrom
claude/4197-team-security-suite

Conversation

@DoodleBears

@DoodleBears DoodleBears commented Nov 16, 2026 •

Copy link
Copy Markdown
Contributor

Why the change

Team mode needs a kept, re-runnable suite. It tries to break the permission model the way a real user would, with real model turns, and it gates every permission-affecting PR.

Closes #4197

Stacked on #4277 (M12 confinement), which the issue lists as a blocker. It also carries the cherry-picked #4298 fix, which this suite found. Retarget to main after both merge.

Special things to note

Change outline

Group Scenario Hard assertion
A Role Policy A0 control, A1 direct / claimed permission / role-play / Memory summary / reversed encoding / relay through own Bot, A2 speaker-policy, A3 Allow+Deny, A4 mixed Group, A5 role revoked mid-conversation Restricted value absent from every reply. Policy items and conflicts read from the Host.
B Files B1 file tools (other member's Memory and account, shared folder, Owner Bot Memory, .credentials.yaml, .. and symlink), B2 Shell script (same targets, grep -r DSH_HOME, find /, write outside, secret env, Host API port), B3 grant requests Sentinel values absent, no file written outside, folder picker lists only own space, member and Owner approval onto foreign folders refused, grants unchanged.
C Shipped boundaries C1 private Bot, private Channel, Human DM, search; C2 model allowlist, deactivation Coded refusals (bot-not-found, remote-access/denied), allowlist not all, deactivated member refused.

Docs:

  • AGENTS.md: the suite is now required for permission-affecting PRs.
  • Updated: verify-deepseekbot skill, features/team-mode.md, docs/agents/development-scripts.md, and the change fragment.

Generated by Claude Code

Member Bot commands run in a bwrap profile that hides DSH_HOME, homes,
/run and other Bots' Memory and Workspaces, binds back only the Session
roots, removes secret environment variables, and joins a private network
namespace that denies the Host loopback, link-local, private and tailnet
ranges. Team mode reports enforced isolation only when both work.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Lpx6qkvJqKRjdg67NfFT7X
The hidden space name started with a NUL byte. SQLite NOCASE comparison
stops at NUL, so every public Bot's space name compared equal and the
second publication on a Host failed the shared-name unique index.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Lpx6qkvJqKRjdg67NfFT7X
scripts/e2e-team-security.mjs sets up roles with Role Policy, members,
public and private Bots, folders and Channels on an isolated Host and
drives real model turns that try to leak restricted figures, reach other
members' folders, the Host data directory and the Host API through file
tools, Shell and grants, and cross private Bots, Channels, DMs,
offboarding and model allowlists. Hard boundaries are asserted on Host
state and tool results. AGENTS.md requires its report for every PR that
changes Team mode permissions.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Lpx6qkvJqKRjdg67NfFT7X
@coderabbitai

coderabbitai Bot commented Nov 16, 2026 •

Copy link
Copy Markdown

Warning

Review limit reached

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Next included review available in 39 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used all 4 included reviews currently available.

Learn how review limits work.

Review configuration:

Run configuration

  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: defaults
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: 09c960cd-92b0-48be-85ff-796b0011f66d
Commits

Reviewing files that changed from the base of the PR and between 3b4dce3 and 1a38edc.

Files selected for processing: 8

  • .agents/skills/verify-deepseekbot/SKILL.md
  • .agents/skills/verify-deepseekbot/features/README.md
  • .agents/skills/verify-deepseekbot/features/team-mode.md
  • AGENTS.md
  • changes/4197-team-security-suite.md
  • docs/agents/development-scripts.md
  • scripts/e2e-team-security.mjs
  • scripts/verification-inventory.json
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

# Conflicts:
#	packages/core/test/team-public-bot-memory.test.ts
Base automatically changed from claude/project-thread-oeczgu to main November 16, 2026 15:28
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Lpx6qkvJqKRjdg67NfFT7X
@DoodleBears
DoodleBears marked this pull request as ready for review November 17, 2026 02:15
@DoodleBears
DoodleBears merged commit 148fb85 into main Nov 17, 2026
15 checks passed
@DoodleBears
DoodleBears deleted the claude/4197-team-security-suite branch November 17, 2026 02:15
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Team mode M15: security scenario suite with real LLM turns

2 participants