The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external content including document titles, highlights, and user notes retrieved via the Readwise API or CLI, as well as a local reader_persona.md file. This data ingestion surface could be exploited if malicious instructions are embedded within the synced reading material or annotations.
Ingestion points: Readwise document data (Step 1 & 2), highlight notes (Step 3), and reader_persona.md (Setup).
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious content within the fetched data.
Capability inventory: The skill is restricted to read-only operations using vendor-specific Readwise tools and local file reading; it does not possess capabilities for file writing, system modification, or external network exfiltration.
Sanitization: No specific sanitization or filtering logic is defined for the external content before it is interpolated into the summary prompt.
[COMMAND_EXECUTION]: The instructions specify the use of the readwise CLI as a fallback mechanism for retrieving data. This involves executing system commands such as readwise list, readwise read <id>, and readwise highlights <id>. These are official vendor tools for the Readwise service.

reader-recap