The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires a CLI binary for core functionality and provides installation instructions via npx -y @mvanhorn/printing-press-library and manual binary downloads from GitHub. These resources are provided by the skill author and are consistent with the tool's intended setup process.
[COMMAND_EXECUTION]: The skill operates by driving the printing-press-oauth2-pp-cli command-line tool to interact with the target API and manage local state, such as authentication tokens and query history.
[REMOTE_CODE_EXECUTION]: The recommended installation path utilizes npx to download and execute the @mvanhorn/printing-press-library package to provision the necessary binary. This represents a vendor-managed remote code execution pattern for tool distribution.
[PROMPT_INJECTION]: The skill processes data from external API responses and a local learning loop (SQLite database and journal logs), presenting an indirect prompt injection attack surface.
Ingestion points: Untrusted data enters the context via API responses from the configured base_url and local stored learnings in data.db and teach.log.
Boundary markers: None identified; the skill instructions do not mandate the use of delimiters or 'ignore' instructions when the agent processes retrieved data.
Capability inventory: The skill allows shell execution via the agent's bash tools and performs network operations and local filesystem writes through the CLI binary.
Sanitization: The underlying Go code implements structural JSON sanitization and PII masking, but lacks specific mechanisms to sanitize natural language instructions embedded within data.

pp-printing-press-oauth2