The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interacts with the macOS operating system by executing the sori CLI tool for tasks such as starting/stopping recordings and managing audio devices.
[INDIRECT_PROMPT_INJECTION]: The skill reads and processes external data from local metadata files which could potentially be manipulated by other local processes to influence the agent.
Ingestion points: The skill reads meta.json files located within the ~/Sori/recordings directory (SKILL.md).
Boundary markers: The instructions include validation steps, such as checking for the ok: true response and status: done state before reporting success.
Capability inventory: The agent has the ability to execute shell commands via the sori binary and can pass data to other tools for transcription or summarization.
Sanitization: The skill verifies specific status flags in the JSON metadata but does not perform broader sanitization of the file content before reporting it to the user.

sori-recorder