doc-and-modernize
Warn
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: During the 'Modernization mode' feasibility spike, the skill instructs the agent to run project-specific build and test commands such as
make,npm install,go build,cargo, orgradle. If the codebase being analyzed is untrusted, these commands can lead to arbitrary code execution via malicious logic embedded in Makefiles or package manager lifecycle scripts. - [PROMPT_INJECTION]: The skill has a significant exposure to indirect prompt injection. It reads extensive content from the local repository (manifests, documentation, source code, and agent-specific files like
AGENTS.md). Maliciously crafted content in these files could attempt to hijack the agent's behavior or influence the generated architecture and modernization documents. - [REMOTE_CODE_EXECUTION]: The assessment workflow involves using package managers to verify dependency installation. This creates a risk of remote code execution if the repository's configuration files (e.g.,
package.json,requirements.txt) reference malicious external packages that execute code during the installation process. - [DATA_EXFILTRATION]: While the skill includes instructions to redact credentials from Git remote URLs, it allows for 'web/API lookups' as a last-resort fallback. This network access capability could be abused to exfiltrate repository metadata or system facts if the agent is manipulated through indirect prompt injection.
Audit Metadata