The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a Python script scripts/landscape_query.py intended for execution by the agent or user. The script is a read-only client that fetches data from the CNCF Landscape API. It uses standard library modules (urllib.request) and implements strict argument parsing and type checking, minimizing the risk of arbitrary command injection.
[EXTERNAL_DOWNLOADS]: The skill performs outbound HTTPS requests to landscape.cncf.io to retrieve project and member metadata. These requests target well-known, official infrastructure of the Cloud Native Computing Foundation (CNCF). The communication is read-only, limited to JSON payloads, and lacks authentication headers, which mitigates credential exposure risks.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from the CNCF Landscape API. To mitigate potential injection risks from third-party project descriptions or summaries, the skill enforces an 'evidence discipline' and 'decision framework' in SKILL.md and references/decision-framework.md. These instructions mandate that the agent distinguish between 'observed' API data and 'inferred' judgment, and explicitly requires human validation through proof-of-concept steps.
[SAFE]: The skill demonstrates high security maturity by including automated tests (tests/test_landscape_query.py) that verify 'fail-closed' behavior for non-JSON or malformed responses, preventing the agent from processing unexpected content types.

cncf-landscape