The Agent Skills Directory

Dynamic Execution (SAFE): The script quick_validate.py correctly uses yaml.safe_load() to parse user-provided metadata, preventing arbitrary object instantiation and potential code execution vulnerabilities commonly associated with unsafe YAML loaders.
Data Exposure & Exfiltration (SAFE): No network operations (e.g., curl, fetch, requests) or access to sensitive local files (e.g., SSH keys, credentials) were found. The script package_skill.py only performs local zip operations.
Unverifiable Dependencies (SAFE): The scripts rely on standard Python libraries (sys, os, re, pathlib, zipfile) and the widely-used PyYAML library. No external scripts are downloaded or executed from the internet.
Indirect Prompt Injection (SAFE): While the validation script ingests external SKILL.md content, it performs strict schema validation, key-whitelisting, and length checks without passing any content to an execution or interpolation context that could be exploited.

skill-creator