OpenClaw Zulip

Security checks across malware telemetry and agentic risk

Overview

This appears to be a coherent Zulip bridge, but it needs Zulip bot credentials and can monitor, post, and administer Zulip content, so install it only with permissions you intend to grant.

Before installing, create a dedicated Zulip bot with the least privileges needed, keep admin actions disabled unless required, leave mention/allowlist policies restrictive, prefer the standard ClawHub install path, and avoid force-install workarounds unless you have reviewed the source and scanner warning.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Risk analysis

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

#
ASI01: Agent Goal Hijack
Low
What this means

People who can message or mention the bot may influence what the agent sees and how it responds, depending on the configured policies.

Why it was flagged

External Zulip events can enter the agent's context and lead to responses, but the skill discloses mention gating and configurable chat modes.

Skill content
When an event arrives from Zulip, the agent will automatically have context ... By default, the bot only responds to @mentions in streams unless `chatmode` is changed.
Recommendation

Keep default mention gating and restrictive DM/group policies unless you intentionally want broader access.

#
ASI02: Tool Misuse and Exploitation
Medium
What this means

If enabled with a powerful bot account, the agent may be able to edit/delete messages or streams, invite users, or deactivate/reactivate users.

Why it was flagged

The plugin exposes high-impact Zulip mutation/admin capabilities, with an explicit configuration gate for admin actions.

Skill content
deleteZulipMessage, deleteZulipStream, ... deactivateZulipUser, ... updateZulipRealm ... "Admin actions require enableAdminActions: true in Zulip config"
Recommendation

Use a dedicated low-privilege Zulip bot, keep `enableAdminActions` off unless needed, and require human review for destructive administrative tasks.

#
ASI03: Identity and Privilege Abuse
Medium
What this means

The installed plugin can act with the permissions of the configured Zulip bot account.

Why it was flagged

The plugin requires Zulip account credentials and correctly marks the API key as sensitive in its manifest.

Skill content
"channelEnvVars": { "zulip": ["ZULIP_API_KEY", "ZULIP_EMAIL", "ZULIP_URL", "ZULIP_SITE", "ZULIP_REALM"] } ... "apiKey": { "label": "API Key", "sensitive": true }
Recommendation

Create a dedicated bot, scope its Zulip permissions narrowly, prefer environment variables for secrets, and rotate the API key if the machine or config is exposed.

#
ASI09: Human-Agent Trust Exploitation
Low
What this means

A user copying the fallback source-install commands might bypass a safety warning without fully reviewing the local source tree.

Why it was flagged

The optional source-install troubleshooting text could encourage users to override or work around scanner warnings, although the recommended path remains ClawHub installation.

Skill content
security scanner blocks `--link` installs ... This is a false positive ... Remove scripts/ first ... openclaw plugins install ./ --force
Recommendation

Prefer the ClawHub install path; if installing from source, do not use `--force` unless you have reviewed the warning and the code.

#
ASI10: Rogue Agents
Low
What this means

The bot may continue receiving and processing Zulip events in the background while the channel is enabled.

Why it was flagged

The bridge is designed to keep monitoring Zulip and persist queue/deduplication metadata locally, which is expected for a channel plugin.

Skill content
"Persistent Event Polling": Automatically resumes from where it left off using locally-persisted queue metadata. ... "Durable Deduplication"
Recommendation

Disable the Zulip channel or remove its credentials when you no longer want continuous monitoring.