Gcalcli

AdvisoryAudited by Static analysis on Apr 30, 2026.

Overview

No suspicious patterns detected.

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 changed or compromised fork could run code on the user's machine and access calendar data or cached Google credentials used by gcalcli.

Why it was flagged

The main workflow runs a custom GitHub fork from a branch rather than a pinned, reviewed release or commit. That remote code would execute locally and interact with the user's Google Calendar OAuth session.

Skill content
IMPORTANT: Always use the custom fork with attachment support: uvx --from "git+https://github.com/shanemcd/gcalcli@attachments-in-tsv-and-json" ... gcalcli
Recommendation

Use an official or audited gcalcli release where possible, pin any fork to a reviewed commit hash, and disclose the dependency in the install metadata.

What this means

The agent may attempt to read a specific person's or organization's calendar data rather than only the installing user's intended calendar.

Why it was flagged

The skill uses persistent Google OAuth credentials and repeatedly scopes commands to a specific work-calendar address with full event details, without saying users should replace it or confirm access boundaries.

Skill content
Cache credentials for future use ... Request calendar read permissions ... gcalcli agenda --calendar smcdonal@redhat.com --details all --json
Recommendation

Require the user to choose the calendar ID at runtime, avoid hardcoded personal/work calendars, and document the exact OAuth scope and credential storage path.

What this means

Private meeting notes, recordings, chat transcripts, or shared document links could be pulled into the agent's working context or saved locally more broadly than the user intended.

Why it was flagged

The skill is designed to retrieve and export sensitive meeting-derived context, including AI notes, recordings, chats, and Drive links, with a bulk-export workflow and no clear retention or reuse limits.

Skill content
Common attachment types: "Notes by Gemini", "Recording", "Chat" ... Direct link to Google Drive file or Google Doc ... Bulk export ALL Gemini notes from search results
Recommendation

Ask for explicit confirmation before exporting attachments or notes, limit searches to user-specified events/date ranges, and explain where exported files are stored and how they should be handled.

ConcernMedium Confidence
ASI05: Unexpected Code Execution
What this means

The agent could execute local project code that was not reviewed as part of the skill and write sensitive meeting exports to the user's Downloads folder.

Why it was flagged

The skill tells the agent to run an undeclared local helper project from a hardcoded user path to export meeting notes, even though the registry has no install spec or reviewed code for that helper.

Skill content
cd /var/home/shanemcd/github/shanemcd/gcmd
uv run gcmd export "$GEMINI_URL" -o ~/Downloads/
Recommendation

Declare and review any helper tools, avoid hardcoded personal paths, and require user confirmation before running local project code or writing exported files.