The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill is designed to ingest and process untrusted data, including source code, security advisories (CVE/GHSA), and user-provided context. Data is passed to subagents for parallel processing, which is a standard pattern for large-scale analysis but increases the surface area for potential indirect injection from the repository being scanned. The skill documentation explicitly notes that user-provided security context should be treated as untrusted analysis data.
[Local Command Execution]: The skill executes local Python scripts to manage the scan workflow. Scripts like generate_rank_input.py and finalize_scan_contract.py are executed using a configurable Python interpreter. These operations are contained within the skill's plugin directory and are used to resolve repository scopes and finalize scan results.

security-scan