Percept Meetings

PassAudited by VirusTotal on May 11, 2026.

Overview

Type: OpenClaw Skill Name: percept-meetings Version: 1.0.0 The skill bundle appears to serve a legitimate purpose of integrating meeting data. However, the `scripts/meeting-context.sh` file passes user-supplied arguments directly to the `percept` command using `"$@"` in the `search` case. This creates a potential shell injection vulnerability if the `percept` command itself does not properly sanitize its input, allowing a malicious agent or user to execute arbitrary commands. While there's no clear evidence of intentional malicious behavior like data exfiltration or backdoors, this vulnerability elevates the classification to suspicious.

Findings (0)

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.

What this means

A Zoom Server-to-Server OAuth app could let the integration read many users' cloud recordings and account user data, not just one meeting transcript.

Why it was flagged

These are admin-level Zoom scopes that can expose recordings and user lists beyond a single user's selected meetings. The registry metadata also declares no primary credential or required env vars, so this privileged access is under-declared.

Skill content
Add scopes:
   - `recording:read:list_recording_files:admin`
   - `recording:read:list_user_recordings:admin`
   - `user:read:list_users:admin`
Recommendation

Use the least-privileged Zoom app possible, restrict it to the intended account or users, and install only if you are comfortable granting account-level recording access.

What this means

Private conversations may become searchable agent context and could be reused in later tasks if the database is not carefully scoped and managed.

Why it was flagged

The Omi workflow can capture broad ambient conversations and persist them into a searchable database. The artifacts do not describe retention limits, exclusions, encryption, participant consent handling, or when stored context should be reused.

Skill content
No calendar integration needed — it captures everything ambient ... Stored in Percept DB, searchable immediately
Recommendation

Configure capture narrowly, get appropriate consent, protect the database, and define retention or deletion practices before using ambient meeting capture.

What this means

The agent could turn transcript-derived action items into real-world account changes or communications without enough review.

Why it was flagged

The instruction encourages the agent to take external follow-up actions based on meeting transcripts, but it does not specify a confirmation step before sending messages, creating tasks, or scheduling events.

Skill content
Use other OpenClaw tools to execute (send emails, create tasks, schedule follow-ups)
Recommendation

Require explicit user approval before sending emails, creating tasks, scheduling meetings, or making other external changes based on meeting data.

What this means

Installing the external package will run and trust code outside this skill bundle.

Why it was flagged

The main functionality depends on installing external code that is not included in the reviewed artifacts and is not pinned to a specific version. This is expected for this integration but should be verified by the user.

Skill content
pip install getpercept   # or: git clone https://github.com/GetPercept/percept
Recommendation

Review the package and repository, pin a trusted version, and install in a controlled environment if possible.