shannon-ai-pentester
Fail
Audited by Gen Agent Trust Hub on Mar 18, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires cloning an external repository from a source not included in the trusted vendors list.
- Evidence:
git clone https://github.com/KeygraphHQ/shannon.gitinSKILL.md. - [COMMAND_EXECUTION]: The skill instructs the user to execute a local shell script (
./shannon) that is part of the untrusted external repository. - Evidence:
./shannon start URL=https://your-app.example.com REPO=/path/to/your/repoinSKILL.md. - [CREDENTIALS_UNSAFE]: The skill asks the user to provide highly sensitive information, including API keys for multiple AI providers and credentials for the target application being tested.
- Evidence: Instructions to set
ANTHROPIC_API_KEY,AWS_SECRET_ACCESS_KEY,TARGET_PASSWORD, andTARGET_TOTP_SECRETinSKILL.md. - [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to analyze source code and web application responses, which are untrusted data sources that could contain malicious instructions designed to influence the agent's behavior.
- Ingestion points:
REPOpath (source code),URL(application responses). - Boundary markers: None specified in the instructions.
- Capability inventory: Network operations (Nmap, exploits), file system access (reading repo), and subprocess execution (
./shannon). - Sanitization: Not explicitly mentioned in the skill definition.
Recommendations
- AI detected serious security threats
Audit Metadata