The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the download and installation of packages from the PRPM registry to enhance agent capabilities.
Evidence: The skill uses prpm install <package-name> --as claude to fetch external content. While it focuses on vendor packages, it explicitly supports community-authored packages like @sanjeed5/github-actions.
[COMMAND_EXECUTION]: The skill executes shell commands to interact with a remote package registry.
Evidence: The workflow utilizes prpm ai-search, prpm search, and prpm install to manage external dependencies.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection from metadata contained within external package search results.
Ingestion points: The agent reads and displays package names, descriptions, and author information returned by the prpm search and prpm ai-search commands.
Boundary markers: There are no instructions providing delimiters or markers to help the agent distinguish between trusted instructions and potentially malicious content embedded in package descriptions.
Capability inventory: The skill has the ability to execute shell commands via the prpm tool and load new logic/knowledge into the agent context.
Sanitization: The content retrieved from the registry search is presented to the user and the agent without evidence of filtering or sanitization.

self-improving