The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests Gherkin scenario facts from the context to generate and execute Java test code. An attacker providing malicious 'facts' could influence the agent to write and execute arbitrary code during the verification phase.
Ingestion points: Gherkin scenario facts provided within the agent's conversation context (referenced in SKILL.md and the reference file).
Boundary markers: The skill explicitly instructs the agent to ignore raw .feature files and only accept 'maintainer-authored or maintainer-sanitized' facts, though this is a soft constraint.
Capability inventory: The agent is instructed to write Java files and execute shell commands (./mvnw compile, ./mvnw clean verify) which provides a path from data ingestion to code execution.
Sanitization: The skill lacks automated sanitization of the input facts, relying entirely on the instruction to only process 'sanitized' data.
[COMMAND_EXECUTION]: The skill workflow involves executing build and verification commands (./mvnw compile, mvn compile, ./mvnw clean verify, mvn clean verify). While these are standard for Java development, they represent the execution of local shell commands within the agent's environment.

323-frameworks-spring-boot-testing-acceptance-tests