The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the agent to install dependencies from vendor-owned sources, specifically using npx to fetch @mvanhorn/printing-press-library and go install to fetch code from github.com/mvanhorn/printing-press-library.
[COMMAND_EXECUTION]: The skill is centered around executing the opensky-pp-cli binary. It instructs the agent to use various subcommands for flight data retrieval, profile management, and interacting with a local learning database.
[DATA_EXFILTRATION]: The CLI supports a --deliver webhook:<url> flag. This feature enables the agent to send the results of its commands (which may include flight data or local configuration) to an external URL via an HTTP POST request.
[REMOTE_CODE_EXECUTION]: Through its 'Learning Loop' and 'Playbook' systems, the skill requires the agent to follow a decision tree where it may execute 'trial commands' or 'playbook steps' synthesized by the CLI at runtime based on previously stored or recalled data.
[PROMPT_INJECTION]: The skill implements an architecture that processes untrusted data from both a public API and a local history of natural language queries.
Ingestion points: External data is ingested via the OpenSky Network API, local SQLite data stores, and input files passed to the recall and teach commands.
Boundary markers: The instructions explicitly warn the agent against interpolating user content into shell command lines, providing a mitigation strategy to treat user input as opaque data.
Capability inventory: The tool possesses the ability to execute shell commands, perform network operations (API calls and webhooks), and manage local file state.
Sanitization: The skill mandates the use of file-based input (e.g., --query-file "$tmpfile") to ensure that user-provided strings are not interpreted as shell commands.

pp-opensky