The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes the implementation of service mocking tools (WireMock and Mountebank) which ingest and process untrusted network data including HTTP, TCP, and SMTP traffic. The execution environment suggested for these tools includes broad capabilities such as shell command execution (mvn, npm, curl, docker) and network access. While the instructions include explicit warnings to avoid recording PII and to scrub data before storage, the ingestion of external data combined with these tool capabilities creates a potential surface for indirect prompt injection if the agent is tasked with processing or acting upon recorded traffic.
[DYNAMIC_EXECUTION]: The documentation for Mountebank (in references/mountebank.md) describes the allowInjection: true configuration, which enables the execution of arbitrary JavaScript code for custom predicates and responses. This represents a dynamic execution capability where code is assembled and run at runtime. The skill correctly categorizes this as an anti-pattern for production-adjacent environments and warns that it allows arbitrary code execution.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to download and install external dependencies from well-known sources, including the @mbtest/mountebank package from NPM and the bbyars/mountebank image from Docker Hub. These are recognized community tools for the tasks described.
[COMMAND_EXECUTION]: The skill provides numerous examples of shell command execution, including building projects with Maven, starting background processes with npx, managing containers with docker, and using curl to interact with control APIs. These are standard developer operations but define the privilege level of the environment.

wiremock-stubs