shannon
Warn
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill downloads the Shannon pentesting framework via 'git clone' from 'https://github.com/KeygraphHQ/shannon.git' and subsequently executes the './shannon' script. This sequence of downloading and executing remote code is the core operational mechanism of the skill.
- [COMMAND_EXECUTION]: The skill extensively uses the Bash tool to manage the local environment, including creating symlinks for source code, pulling Git updates, and orchestrating Docker containers to run pentest agents.
- [EXTERNAL_DOWNLOADS]: Fetches the Shannon framework and updates from a non-whitelisted GitHub repository (KeygraphHQ) to establish the testing environment.
- [SAFE]: The skill possesses a significant attack surface for indirect prompt injection as it ingests target application responses and source code. However, this risk is mitigated by explicit instructions for the agent to perform authorization checks and verify that the target is not a production system.
Audit Metadata