The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to run local Python scripts (scripts/init_skill.py and scripts/package_skill.py) to manage the skill development lifecycle. These scripts interact with the host file system to create project structures and package assets into ZIP archives.
[PRIVILEGE_ESCALATION]: The scripts/init_skill.py script uses chmod 0o755 to grant execution permissions to a boilerplate Python script (scripts/example.py) that it generates at runtime. Modification of file permissions on dynamically created content is a sensitive capability, though it is used here to facilitate tool development.
[DYNAMIC_EXECUTION]: The scripts/init_skill.py script performs low-risk dynamic code generation by writing a predefined Python template into a new file (scripts/example.py) within the user-specified directory.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it accepts and processes user-defined inputs, such as skill names and directory paths, which are used in file system operations.
Ingestion points: Command-line arguments provided to scripts/init_skill.py and scripts/package_skill.py as described in SKILL.md.
Boundary markers: Absent. The scripts do not implement specific delimiters or warnings for the agent regarding the handling of these inputs.
Capability inventory: The skill scripts utilize capabilities for directory creation (mkdir), file writing (write_text), permission modification (chmod), and archiving (zipfile).
Sanitization: While scripts/quick_validate.py contains regex validation for skill names (enforcing hyphen-case), these checks are not integrated into the init_skill.py workflow before it executes directory creation and file writes, creating a theoretical risk of path traversal if malicious inputs are supplied.

skill-creator