The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The scripts/init_skill.py tool creates a new script file and immediately grants it executable permissions (chmod 755). This represents a privilege escalation pattern where the system automatically upgrades file permissions for generated content.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process other skill files (SKILL.md) for merging or upgrading tasks. This process creates a potential surface for indirect prompt injection, as the agent reads instructions from external markdown files without enforced boundary markers or content sanitization.
[DYNAMIC_EXECUTION]: The skill uses scripts/init_skill.py to generate new Python scripts and markdown documentation from hardcoded templates at runtime. This dynamic generation of executable or instructional content is a known pattern for creating automated workflows.
[COMMAND_EXECUTION]: The workflow defined in the skill involves the execution of local Python scripts to validate and package skills. This indicates that the skill requires the agent to have permissions to run commands on the local operating system.

skill-writing