The Agent Skills Directory

[COMMAND_EXECUTION]: The core functionality of the skill involves looking up and running shell scripts or multi-step shell pipelines from local and global directories. The instructions explicitly direct the agent to 'Run a matching entry with the arguments the task needs'.
[DYNAMIC_EXECUTION]: The skill instructs the agent to write new executable scripts when existing ones do not fit the task. It specifically requires the agent to 'Run that example once before indexing', which involves executing newly generated code immediately at runtime.
[PERSISTENCE]: The skill utilizes a global scriptbook directory located at ~/.agents/scripts/. Scripts saved to this location persist across different terminal sessions, repositories, and agent interactions, creating a permanent execution surface.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection if the script directories are poisoned with malicious content.
Ingestion points: The agent reads the contents of INDEX.md and individual script files from both project-specific (.agents/scripts/) and global (~/.agents/scripts/) locations.
Boundary markers: None. The agent is instructed to read the index files and script headers to determine suitability and then execute the scripts directly.
Capability inventory: The agent has the capability to write files (saving new scripts) and execute shell commands (running existing or newly created scripts).
Sanitization: There are no instructions for sanitizing or validating the contents of the scripts before execution.

scriptbook