The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process project-level configuration files such as package.json, pnpm-lock.yaml, and aube-workspace.yaml. These files are external, potentially untrusted data sources that could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The skill reads package.json, various lockfiles (pnpm-lock.yaml, package-lock.json, yarn.lock, bun.lock, aube-lock.yaml), and workspace configurations from the file system.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat content from these files as untrusted data or to ignore embedded instructions.
Capability inventory: The skill utilizes commands like aube install, aube run, aube exec, and aube dlx which are capable of executing scripts and binaries defined within the ingested configuration files.
Sanitization: No sanitization or validation logic is specified for how the agent should handle the content of these files before performing actions based on them.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the Aube tool from external repositories and registries, including @endevco/aube via npm and the endevco/tap/aube tap via Homebrew. It also references documentation and benchmarks hosted at aube.en.dev.
[COMMAND_EXECUTION]: The core functionality of the skill involves documenting the use of the aube command-line interface and its subcommands. These commands perform file system operations (managing node_modules and global stores) and network operations (fetching packages from registries).

aube-package-manager