The Agent Skills Directory

[DATA_EXFILTRATION]: The skill documents the --deliver webhook:<url> flag, which enables the output of any command to be sent to an arbitrary external URL. This creates a significant risk of exfiltration for sensitive information retrieved by the CLI, such as account subscriptions, favorites, notifications, and workspace metadata.
[DYNAMIC_EXECUTION]: The skill implements a 'Playbook' system that retrieves and executes sequences of commands (steps) from a local database. These steps are executed dynamically with slot substitutions ({slot}), introducing risks related to runtime command assembly and potential execution of malicious choreography.
[INDIRECT_PROMPT_INJECTION]: The skill features an autonomous 'learning loop' that processes untrusted marketplace data to derive future behavior.
Ingestion points: Data from the RapidAPI marketplace is ingested via sync and search commands and stored in a local SQLite database.
Boundary markers: The instructions do not provide delimiters or safety warnings to ensure the agent ignores potential instructions embedded within marketplace metadata or API descriptions.
Capability inventory: The skill is authorized to execute bash commands, perform network operations via GraphQL and webhooks, and manage local configuration and credentials.
Sanitization: No sanitization or validation mechanisms are described for the marketplace content that is used to synthesize playbook_candidates or results in the learning store.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of external binaries from remote sources, specifically targeting the github.com/mvanhorn/printing-press-library repository and the @mvanhorn/printing-press-library NPM package. These resources are provided by the identified vendor.

pp-rapidapi