labarchive-integration
Fail
Audited by Gen Agent Trust Hub on Feb 15, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
- EXTERNAL_DOWNLOADS (HIGH): The skill instructs users to install the
labarchives-pypackage directly from an untrusted personal GitHub repository (github.com/mcmero/labarchives-py) instead of a verified organization or the official PyPI registry. - REMOTE_CODE_EXECUTION (HIGH): The skill executes code from the unverified
labarchivespylibrary. Since the installation source is untrusted, this allows for arbitrary code execution if the repository contains malicious logic or is compromised. - CREDENTIALS_UNSAFE (HIGH): The
scripts/setup_config.pyscript prompts users for high-privilege credentials, including institutional access passwords and user passwords, and stores them in plain text inconfig.yaml. This creates a local target for credential theft and potential exposure if an AI agent reads or leaks the file content. - PROMPT_INJECTION (MEDIUM): The skill is vulnerable to indirect prompt injection (Category 8). It ingests untrusted external data (notebook entries, comments, and file attachments) and stores them locally via
scripts/notebook_operations.py. If a downstream agent or process analyzes these backups, malicious instructions embedded in the LabArchives content could influence subsequent agent behavior.
Recommendations
- AI detected serious security threats
Audit Metadata