signed-audit-trails-recipe
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and initialize the
protect-mcpand@veritasacta/verifypackages from the NPM registry usingnpx. These tools are used for cryptographic key generation and receipt verification. - [COMMAND_EXECUTION]: The instructions involve executing shell commands to manage sensitive files, such as setting restrictive file permissions (
chmod 600 ./protect-mcp.key) and updating.gitignoreto prevent credential exposure in version control. - [INDIRECT_PROMPT_INJECTION]: The skill explicitly addresses risks associated with tool usage by implementing a Cedar policy layer. This provides a robust mitigation against malicious instructions embedded in tool inputs by enforcing a strict allowlist for shell commands and file access paths.
- [SAFE]: The skill follows security best practices, such as discouraging the commitment of private keys to repositories and providing clear guidance on key rotation and CI/CD security (e.g., using branch-limited environment secrets).
Audit Metadata