The Agent Skills Directory

Command Execution: The skill invokes various system commands and package managers including pip, npm, cargo, mvn, gradlew, and trivy. While these are necessary for the skill's purpose, executing project-local scripts like ./gradlew is a security consideration as it runs code defined within the repository being audited.
External Tool Installation: The skill may install missing audit tools (e.g., pip install pip-licenses, npm install -g license-checker) if they are not present on the system. These downloads originate from well-known package registries.
Dynamic Script Generation: For Gradle projects, the skill generates a temporary initialization script (/tmp/license-report.init.gradle) to apply the license-report plugin dynamically. This is a standard method for non-invasive analysis but involves runtime code generation and execution.
Indirect Prompt Injection Surface: The skill ingests untrusted data from external package registries (package descriptions and license metadata). It includes explicit instructions to treat this metadata as data rather than directives, which serves as a mitigation for potential injection attempts hidden in dependency information.
Pre-flight Module Execution: The skill runs a pre-flight check using python3 -m setup_preflight with a modified PYTHONPATH. This involves executing a local module to verify project readiness before the audit begins.

dependency-license-audit