The Agent Skills Directory

[COMMAND_EXECUTION]: The test scripts in the scripts/ directory (test_audit_skill_metadata.py, test_build_skill_graph.py, and test_validate_skill.py) use subprocess.run to execute the main validator scripts using the current Python interpreter (sys.executable). This is a benign and standard practice for internal regression testing.
[EXTERNAL_DOWNLOADS]: The scripts/validate_skill.py utility uses urllib.request to perform network checks on external URLs defined in data/sources.json. This functionality is intended for link validation and is only triggered when the --check-urls command-line flag is explicitly provided.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves parsing and analyzing other skill bundles (including SKILL.md frontmatter and openai.yaml files). While this processes external data, the analysis is limited to regex-based static checks and does not evaluate or execute the content in a way that would facilitate prompt injection attacks against the host agent.
[SAFE]: The data/sources.json file contains a curated list of references to official documentation and community tools from trusted organizations such as Anthropic, Microsoft, OpenAI, and well-known GitHub repositories. These references are documented neutrally and used for technical verification.

agents-skills