The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs several shell operations to manage Elixir projects, including initializing projects with mix new, fetching dependencies with mix deps.get, compiling code with mix compile, and running tests via mix test. It also uses filesystem commands like cp to manage OpenAPI specifications and grep to locate existing modules.
[EXTERNAL_DOWNLOADS]: The skill configures projects to download the can_opener library from the author's repository at https://github.com/agoodway/can_opener.git. It also supports fetching OpenAPI specifications from user-provided URLs.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from an OpenAPI JSON file (provided via a path or URL) and uses it to generate Elixir function names, schemas, and logic. This creates a surface for indirect prompt injection, as a malicious specification could attempt to influence the agent's behavior during the scaffolding, code discovery, or verification steps. Evidence: Ingestion of openapi.json in SKILL.md and subsequent execution of mix run and mix test on the generated output.
[DYNAMIC_EXECUTION]: The skill generates Elixir source code (lib/<package_name>.ex and test/<package_name>_test.exs) based on the provided OpenAPI spec. It then invokes the Elixir compiler and runtime to build and verify the generated code.

can-opener