The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's 'apply' functionality is designed to execute a list of preCommands stored within a JSON template file. While the documentation provides benign examples like git status, the skill does not implement any restrictions, allowing the execution of any arbitrary shell command accessible to the agent.
[DYNAMIC_EXECUTION]: The skill uses the 'Bash' tool to execute instructions defined dynamically at runtime from local configuration files (~/.claude-context-optimizer/templates/*.json). This behavior creates a risk where modifying a template file could lead to the agent performing unintended actions.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by processing external data (the template files) that controls agent actions.
Ingestion points: The skill reads JSON template files from the user's home directory (~/.claude-context-optimizer/templates/) to determine its next steps.
Boundary markers: There are no explicit boundary markers or instructions to ignore potential commands embedded within the data fields of the template files.
Capability inventory: The skill utilizes Bash for command execution, Read for file access, Write for template creation, and Glob for file searching.
Sanitization: The skill does not perform any validation or sanitization of the preCommands array before passing it to the shell for execution.

cco-templates