The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/trace_vulns.py script executes external commands such as npm ls and pipdeptree using subprocess.run(). These invocations use list-based arguments rather than shell strings, which is a secure practice that prevents command injection vulnerabilities.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, creating a potential surface for indirect prompt injection where malicious metadata could influence the agent's behavior. Ingestion points: The scripts/trace_vulns.py script reads JSON data via the --audit-input parameter and parses local project files like requirements.txt. Boundary markers: The skill does not use specific delimiters or warnings to isolate ingested data from agent instructions. Capability inventory: The skill has access to tools including Bash (for npm, pip, and python3) and file system read capabilities. Sanitization: The script performs standard JSON and text parsing but does not sanitize or filter the content of strings extracted from the target files.
[EXTERNAL_DOWNLOADS]: The documentation in references/PLAYBOOK.md recommends installing reputable third-party security utilities, such as @cyclonedx/cyclonedx-npm, cyclonedx-bom, and syft. These references target well-known organizations and official registries.

tracing-transitive-vulnerabilities