The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The scripts/init_skill.py script automatically modifies file permissions on generated template files using chmod(0o755). This programmatic elevation of privileges on dynamically created files is a recognized security pattern.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of other skills, which involves ingesting user-provided data into instructions and scripts that will be executed later.
Ingestion points: User input for skill names in the initialization script and manual content entry in the generated markdown and resource files.
Boundary markers: Generated templates do not include specific delimiters or instructions for the agent to ignore potentially malicious embedded content.
Capability inventory: The skill is granted file system access through Bash tools (cat, echo, ls) and utilizes Python's pathlib for writing and permission modifications.
Sanitization: The implementation provides basic regex-based name normalization and simple character filtering in descriptions, but lacks comprehensive sanitization for the core instructional content.
[COMMAND_EXECUTION]: The skill relies on allowed Bash tools to read, write, and list files as part of its initialization and packaging routines.
[DYNAMIC_EXECUTION]: The provided instructions guide the agent to write scripts and then verify them by running the resulting code, which constitutes a dynamic execution workflow.

skill-creator