The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands such as ls, which, and openpaw to manage system state, verify installed components, and launch a local web dashboard.- [EXTERNAL_DOWNLOADS]: The skill references the openpaw add command for installing additional capabilities from the project's repository.- [DATA_EXFILTRATION]: The skill reads and writes sensitive data, including personality definitions and long-term memory, to files located in the user's home directory at ~/.claude/SOUL.md and ~/.claude/memory/.- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted data from ~/.claude/SOUL.md and ~/.claude/memory/MEMORY.md to define its personality and behavior.
Ingestion points: Reads from ~/.claude/SOUL.md and ~/.claude/memory/MEMORY.md at session start.
Boundary markers: None present; instructions are followed directly from the file content.
Capability inventory: Capability to execute shell commands (ls, which, openpaw) and write to files.
Sanitization: None present; the agent is instructed to adopt the personality defined in the external files.- [PROMPT_INJECTION]: The 'PAW MODE' section contains instructions that attempt to override the agent's standard operating persona and response style.