The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the user's personal activity history, including screen text, audio transcripts, and app logs, which may contain malicious instructions.
Ingestion points: Output from the luci CLI search and transcript commands, as well as markdown files in the ~/.Life directory.
Boundary markers: The instructions lack delimiters or specific warnings to ignore instructions embedded within the retrieved data.
Capability inventory: The skill executes shell commands using the resolved <CLI> path and can write files to the local system (e.g., image <ID> -o <output.jpg>).
Sanitization: No evidence of sanitization or filtering of external content before interpolation into the agent's context.
[COMMAND_EXECUTION]: The skill dynamically resolves and executes a local command-line tool based on platform-specific configuration files.
Evidence: Resolves the path to the luci executable from ~/.luci/cli.json or ~/.luciMicrosoft/cli.json and executes various subcommands for data retrieval.

luci