CVE-2026-92801: cc-connect Auth Bypass Vulnerability

CVE-2026-92801 Overview

CVE-2026-92801 is an authorization bypass vulnerability in cc-connect through version 1.5.0. The application fails to enforce per-user allowlist filtering inside the onCardAction handler that processes Feishu interactive card callbacks. Attackers can dispatch agent commands by triggering card actions in admitted chats. This bypasses the per-user access controls that guard the text message handler. The weakness is classified under CWE-863: Incorrect Authorization.

Critical Impact
Authenticated users in admitted chats can execute agent commands they should not be authorized to run, undermining the confidentiality, integrity, and availability of the connected agent backend.

Affected Products

cc-connect through version 1.5.0
Feishu (Lark) integration component in platform/feishu/feishu.go
Deployments exposing the onCardAction callback handler

Discovery Timeline

2026-09-16 - CVE-2026-92801 published to NVD
2026-09-17 - Last updated in NVD database

Technical Details for CVE-2026-92801

Vulnerability Analysis

The cc-connect project bridges chat platforms with agent backends. On the Feishu platform, user messages flow through a text message handler that consults a per-user allowlist before dispatching agent commands. The parallel onCardAction handler processes interactive card callbacks such as button clicks and form submissions from the same chats. The card action handler omits the allowlist check that the text handler performs. Any authenticated user who can trigger a card action inside an admitted chat can invoke agent commands reserved for allowlisted users.

Root Cause

The root cause is an authorization control that is applied inconsistently across two code paths that dispatch the same privileged operations. The vulnerable logic resides in platform/feishu/feishu.go between lines 661 and 680 of tag v1.5.0. See the GitHub code snippet and the VulnCheck advisory for the exact location and reasoning. The design assumed only allowlisted users would render cards, so the callback path was left unchecked.

Attack Vector

An attacker with membership in an admitted Feishu chat sends or interacts with an interactive card. Triggering the card action generates a callback that cc-connect routes to onCardAction. The handler dispatches the underlying agent command without validating the invoking user against the allowlist. This yields command execution equivalent to a trusted operator. Refer to the cc-connect repository and the issue tracker entry for reproduction details.

Detection Methods for CVE-2026-92801

Indicators of Compromise

Feishu card action callbacks in cc-connect logs where the invoking open_id or union_id is not present in the configured per-user allowlist.
Agent command executions that originate from onCardAction events rather than plain text messages, especially outside normal operator accounts.
Anomalous bursts of card interactions in admitted chats followed by outbound agent activity.

Detection Strategies

Audit cc-connect runtime logs for onCardAction invocations and correlate the user_id field with the sanctioned allowlist.
Add server-side logging that records the pre-dispatch authorization decision for both text and card handlers to expose the missing check.
Compare command dispatch rates between text messages and card actions to detect unusual reliance on the card path.

Monitoring Recommendations

Forward Feishu webhook and cc-connect application logs to a centralized analytics platform for cross-user correlation.
Alert on any agent command executed by a non-allowlisted principal, regardless of the originating handler.
Track deployments of cc-connect versions at or below 1.5.0 across the estate to prioritize remediation.

How to Mitigate CVE-2026-92801

Immediate Actions Required

Restrict Feishu chat membership so only trusted operators can send messages or interact with cards in admitted chats.
Disable or unregister interactive card publishers from cc-connect deployments running version 1.5.0 or earlier until a fixed release is applied.
Rotate agent credentials and API tokens that may have been reachable through unauthorized command dispatch.

Patch Information

At the time of publication, no fixed release is referenced in the NVD entry. Monitor the cc-connect GitHub repository and the tracking issue for a patched version. The fix must add the same per-user allowlist enforcement used by the text message handler to the onCardAction handler in platform/feishu/feishu.go.

Workarounds

Fork cc-connect and add an allowlist check at the top of onCardAction that mirrors the text handler logic before command dispatch.
Place an authorizing proxy in front of the Feishu webhook endpoint that rejects card callbacks from users outside the allowlist.
Limit admitted chats to single-user direct messages with allowlisted operators to eliminate the multi-user exposure.

bash

# Configuration example: restrict Feishu webhook to a vetted allowlist at the reverse proxy
# nginx snippet enforcing that only pre-approved Feishu user IDs reach cc-connect
location /feishu/callback {
    if ($http_x_feishu_user_open_id !~ "^(ou_operator1|ou_operator2)$") {
        return 403;
    }
    proxy_pass http://cc_connect_upstream;
}