The Agent Skills Directory

[SAFE]: The skill implements a structured methodology for skill creation, focusing on discovery, architecture, and validation. It does not exhibit any malicious patterns such as prompt injection, persistence mechanisms, or credential harvesting.
[COMMAND_EXECUTION]: The skill provides a Python validation utility (scripts/validate_skill.py) meant to be executed locally. The script is a static analysis tool that verifies file casing, frontmatter formatting, and structural requirements using standard libraries. It does not perform network operations or dangerous system calls.
[DYNAMIC_EXECUTION]: The included validation script uses the secure yaml.safe_load() method for parsing frontmatter when the PyYAML library is available, preventing potential arbitrary code execution during the parsing phase. It also includes a manual, regex-based fallback parser for environments without PyYAML.
[INDIRECT_PROMPT_INJECTION]: As a tool designed to ingest user requirements and generate instructions, the skill inherently possesses a surface for indirect prompt injection. However, it mitigates this by enforcing structured formatting, explicit delimiters (---), and precise imperative language to create clear boundaries for agent instructions.

skill-architect