The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands such as mkdir -p, echo, and git to manage the contact registry file. The use of user-supplied variables like contact names in command strings (e.g., git commit -m "contacts: add <name>") creates a risk of command injection if the input contains shell metacharacters.
[DATA_EXFILTRATION]: The skill stores and displays absolute local filesystem paths for 'inbox' transports (e.g., /home/peter/Code/peter-oracle/ψ/inbox). This exposes the directory structure of the host machine to the agent's context and potentially to other users if the registry is shared via Git.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes and displays data from an external file (ψ/contacts.json).
Ingestion points: Data is read from ψ/contacts.json and .oracle/contacts.json at runtime.
Boundary markers: No delimiters or warnings are used to separate contact details (like notes or names) from instructions when displaying them to the agent.
Capability inventory: The skill possesses the ability to execute shell commands (git, mkdir, echo) and perform filesystem writes.
Sanitization: No validation or sanitization of the JSON content is performed before processing or displaying the data.

contacts