The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted command files provided by the user or found in specific project directories, which can influence the agent's behavior during the conversion process.
Ingestion points: The skill instructs the agent to read source command files from ./commands/ or ~/.claude/commands/ using the Read or Bash tools.
Boundary markers: The instructions do not provide explicit boundary markers or directives to ignore instructions embedded within the source command content being processed.
Capability inventory: The skill environment has access to Read, Write, and Bash tools, which are used to transform files and execute validation scripts.
Sanitization: The conversion process involves direct text transformation of user-controlled markdown content without sanitization or escaping of the instructions contained therein.
[COMMAND_EXECUTION]: The skill utilizes an automated hook to execute a local Python script whenever the Write tool is invoked.
Evidence: The SKILL.md frontmatter defines a PostToolUse hook for the Write tool that executes python3 ~/.claude/skills/converting-commands-to-skills/scripts/validate-frontmatter.py.
Description: While the script performs benign validation of the skill's YAML frontmatter, it represents an automated shell command execution path triggered by standard tool operations.

converting-commands-to-skills