The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a custom command-line interface readwise-custom to interact with the Readwise API. The tool supports pruning operations with filters for age, location, and tags. All destructive actions are explicitly gated behind a user-confirmed dry-run process.- [INDIRECT_PROMPT_INJECTION]: The skill processes document metadata (titles, tags, categories) from the Readwise Reader service, which serves as the ingestion point. While this introduces an indirect prompt injection surface where external document content could influence agent behavior, the risk is addressed by the mandatory evidence chain: 1) Ingestion point: Readwise API via the CLI tool. 2) Boundary markers: The agent is strictly instructed to display the candidate list to the user first. 3) Capability inventory: The skill is limited to document deletion via readwise-custom. 4) Sanitization: The agent is tasked with manual review via the user.

readwise-prune