The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands including git rev-parse, date, mkdir, git add, and git commit. These are used to establish the project root and manage the lifecycle of the contact registry file.
[PROMPT_INJECTION]: The skill processes and displays untrusted data from ψ/contacts.json. Metadata fields like notes and name are interpolated into the agent's context without explicit boundary markers or sanitization, creating a surface for indirect prompt injection.
Ingestion points: Reads from ψ/contacts.json in listing, showing, and removal modes.
Boundary markers: Absent; data is displayed directly in a table or list format.
Capability inventory: The agent can write to the filesystem via jq or the Write tool and perform Git operations (git add, git commit).
Sanitization: While jq is recommended for updates (structured data), there is no specified validation or escaping for the data when it is read back into the prompt.
[DATA_EXPOSURE]: The skill encourages the storage of absolute local filesystem paths (e.g., /home/peter/Code/peter-oracle/ψ/inbox) in the inbox field. As the skill description notes the file is 'committable and shareable,' this practice can lead to the exposure of local system usernames and directory structures if the repository is pushed to a public or shared remote.

contacts