The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a dedicated script, scripts/run_host_validation.py, which executes local validation commands. This feature is mitigated by several security controls: it uses subprocess.run with shell=False to prevent shell injection, performs redaction of potential secrets in command output using the redact_text function, and enforces a project policy that disables high-risk actions like network access and package installation by default.
[DATA_EXFILTRATION]: Protective measures are integrated into the asset packaging logic. Both scripts/package_skill.py and scripts/package_asset_bundle.py perform a structural audit that rejects symbolic links, preventing path traversal attacks. Furthermore, scripts/scan_secrets.py automatically scans files for patterns resembling API keys or private keys and redacts them, blocking the packaging process if unresolved secrets are found.
[REMOTE_CODE_EXECUTION]: The scripts/validators/skill.py script uses the Python compile() function for syntax verification. This is a deterministic check to ensure code quality within the asset scripts and does not involve the execution of the code's logic. All imported content is treated as untrusted data in accordance with the skill's security provenance guidelines.
[EXTERNAL_DOWNLOADS]: The skill documentation (SOURCES.md, SOURCE-URLS.txt) references official resources from well-known and trusted organizations such as Microsoft, OpenAI, Anthropic, and GitHub. These references are used solely for informational purposes and to define validation schemas, with no automated execution of external remote scripts detected.

mols-agent-asset-studio